Home » Seminari

News, eventi e seminari

Salta il menu di secondo livello


DNA combinatorial messages and Epigenomics: The case of chromatin organization and nucleosome occupancy in eukaryotic genomes

Dove / Data e Ora: 1C150, Torre Archimede, il giorno 03/10/2019 alle ore 14:30

Relatore: Professor Raffaele Giancarlo (Universita' degli Studi di Palermo)

Abstract: Epigenomics is the study of modifications on the genetic material of a cell that do not depend on changes in the DNA sequence, since those latter involve specific proteins around which DNA wraps. The end result is that epigenomic changes have a fundamental role in the proper working of each cell in Eukaryotic organisms. A particularly important part of Epigenomics concentrates on the study of chromatin, that is, a fiber composed of a DNA-protein complex and very characterizing of Eukaryotes. Understanding how chromatin is assembled and how it changes is fundamental for Biology. In more than thirty years of research in this area, Mathematics and Theoretical Computer Science have gained a prominent role, in terms of modeling and mining, regarding in particular the so-called 10nm fiber. Starting from some very basic notions of Biology, we briefly illustrate the recent advances obtained via laboratory experiments on the organization and dynamics of chromatin. Then, we mainly concentrate our attention on the contributions given by Combinatorial and Informational Methodologies, that are at the hearth of Mathematics and Theoretical Computer Science, to the understanding of mechanisms determining the 10nm fiber. We conclude highlighting several directions of investigation that are perceived as important and where Mathematics and Theoretical Computer Science can provide high impact results.

CV Relatore: - - -

Torna su ▲

Learning Fair and Transferable Representations

Dove / Data e Ora: 1BC50, il giorno 24/06/2019 alle ore 16:30

Relatore: Luca Oneto (University of Pisa)

Abstract: Developing learning methods which do not discriminate subgroups in the population is a central goal of algorithmic fairness. One way to reach this goal is by modifying the data representation in order to meet certain fairness constraints. In this work we measure fairness according to demographic parity. This requires the probability of the possible model decisions to be independent of the sensitive information. We argue that the goal of imposing demographic parity can be substantially facilitated within a multitask learning setting. We leverage task similarities by encouraging a shared fair representation across the tasks via low rank matrix factorization. We derive learning bounds establishing that the learned representation transfers well to novel tasks both in terms of prediction performance and fairness metrics. We present experiments on three real world datasets, showing that the proposed method outperforms state-of-the-art approaches by a significant margin.

CV Relatore: Luca Oneto was born in Rapallo, Italy in 1986. He received his BSc and MSc in Electronic Engineering at the University of Genoa, Italy respectively in 2008 and 2010. In 2014 he received his PhD from the same university in the School of Sciences and Technologies for Knowledge and Information Retrieval with the thesis ``Learning Based On Empirical Data''. In 2017 he obtained the Italian National Scientific Qualification for the role of Associate Professor in Computer Engineering and in 2018 he obtained the one in Computer Science He worked as Assistant Professor in Computer Engineering at University of Genoa from 2016 to 2019. In 2018 he was co-funder of the spin-off ZenaByte s.r.l. In 2019 he obtained the Italian National Scientific Qualification for the role of Full Professor in Computer Science and Computer Engineering. He is currently Associate Professor in Computer Science at University of Pisa. His first main topic of research is the Statistical Learning Theory with particular focus on the theoretical aspects of the problems of (Semi) Supervised Model Selection and Error Estimation. His second main topic of research is Data Science with particular reference to the solution of real world problems by exploiting and improving the most recent Learning Algorithms and Theoretical Results in the fields of Machine Learning and Data Mining.

Torna su ▲

CSI NN: Reverse Engineering of Neural Network Architectures Through Electromagnetic Side Channel

Dove / Data e Ora: 2BC60 (DM - Torre Archimede), il giorno 30/05/2019 alle ore 11:00

Relatore: Stjepan Picek (TU Delft, The Netherlands)

Abstract: Machine learning has become mainstream across industries. Numerous examples prove the validity of it for security applications. In this work, we investigate how to reverse engineer a neural network by using side-channel information such as timing and electromagnetic (EM) emanations. To this end, we consider multilayer perceptrons and convolutional neural networks as the machine learning architectures of choice and assume a non-invasive and passive attacker capable of measuring those kinds of leakages. We conduct all experiments on real data and commonly used neural network architectures in order to properly assess the applicability and extendability of those attacks. Practical results are shown on an ARM Cortex-M3 microcontroller, which is a platform often used in pervasive applications using neural networks such as wearables, surveillance cameras, etc. Our experiments show that a side-channel attacker is capable of obtaining the following information: the activation functions used in the architecture, the number of layers and neurons in the layers, the number of output classes, and weights in the neural network. Thus, the attacker can effectively reverse engineer the network using merely side-channel information such as timing or EM.

CV Relatore: Stjepan Picek is an assistant professor in the Cybersecurity group at TU Delft, The Netherlands. His research interests are security/cryptography, machine learning, and evolutionary computation. Prior to the assistant professor position, Stjepan was a postdoctoral researcher at ALFA group, MIT, USA. Before that, he was a postdoctoral researcher at KU Leuven, Belgium as a part of the Computer Security and Industrial Cryptography (COSIC) group. Stjepan finished his PhD in 2015 with a topic on cryptology and evolutionary computation techniques. Stjepan also has several years of experience working in industry and government. Up to now, Stjepan gave more than 10 invited talks at conferences and summer schools and published more than 70 refereed papers in both evolutionary computation and cryptography journals and conferences. Stjepan is a member of the organization committee for International Summer School in Cryptography and president of the Croatian IEEE CIS Chapter. He is a general co-chair for Eurocrypt 2020, program committee member and reviewer for a number of conferences and journals, and a member of several professional societies.

Torna su ▲

No Free Charge Theorem 2.0: How to Steal Private Information from a Mobile Device Using a Powerbank

Dove / Data e Ora: 1A150, il giorno 17/04/2019 alle ore 16:30

Relatore: Dr. Riccardo Spolaor (University of Oxford (UK))

Abstract: Thanks to their omnipresence and multi-purposeness, users rely on smartphones to execute in few touches a wide range of privacy-related operation, such as accessing bank accounts, checking emails, or transferring money. While not long time ago users were seeking constant Internet connection (e.g., via free Wi-Fi hotspot), now they also look for energy sources to recharge their smartphones' battery, due to the use of more energy-draining apps (e.g., Pokémon Go). This phenomenon has led to the diffusion of free charging stations in public places and the marketing of portable batteries a.k.a. powerbanks. Despite the preventive measures implemented by Android's developers in order to prevent data transfer via USB cable (i.e., ''Charging only'' mode), we are able to exploit a hidden communication channel which leverages only the electrical current provided for charging the smartphone. On one side, a malicious app (which can be disguised as a legitimate, clean app) installed on the victim's phone, which only requires wakelock permission (i.e., to wake up the phone when it is idle), remains silent until the device is plugged to a USB port and left unattended. Then, such app begins transmitting sensitive data coded in energy consumption peaks. On the other side, the energy provider (e.g., powerbank) is able to measure such peaks and then reconstruct the transmitted information. All this happens without the malicious app's access to Internet or other permissions, except for the information that it wants to exfiltrate.

CV Relatore: Riccardo Spolaor is a Research Associate at the University of Oxford. He obtained his Ph.D. in Brain, Mind, and Computer Science at the University of Padua, Italy, in 2018. He obtained his Master's Degree in Computer Science in 2014 from the same university, with a thesis about smartphone privacy attack inferring user actions via traffic analysis. In November 2014, he started his Ph.D. under the supervision of Prof. Mauro Conti. He has been a Visiting Ph.D. Student at Radboud University (2015), Ruhr-Universitat Bochum (2016), and University of Oxford (2016, 2017, and 2018). His main research interests are privacy and security issues on mobile devices. In particular, he applies machine learning techniques to infer user information relying on side-channel analysis. Most of the research that he carried out up to now is about the application of machine learning classifiers to network traffic and energy consumption traces.

Torna su ▲

Learning for structures: advances for efficient recursive and deep approaches

Dove / Data e Ora: 1BC45 (DM - Torre Archimede), il giorno 04/04/2019 alle ore 17:00

Relatore: Prof. Alessio Micheli (Computational Intelligence and Machine Learning Group (CIML), University of Pisa)

Abstract: Moving to Structured Domains (SD - sequences, trees and graphs data) is a mainstream in the current evolution of Neural Networks and Machine Learning, which is rooted in foundational works developed in the last 20 years. Recent advancements which still is accompanying the current “deep learning revolution” are often at the price of high computation cost, empathizing the need of efficient approaches. We will use results from my research group to show how we can move toward deep and efficient approaches for SD (both for sequences, trees and graphs) and, more in general, to show the many potential benefits in the extension and analysis of the impact of deep approaches for SD.

CV Relatore: Prof. Dr. Alessio Micheli is Associate Professor at the Department of Computer Science of the University of Pisa, where he is the head of the Computational Intelligence and Machine Learning Group (CIML). He is the national coordinator of the ''Italian Working group on Machine Learning and Data Mining'' of AI*IA. His research interests include machine learning, neural networks, deep learning, sequence and structured domains learning, recurrent and recursive neural networks, reservoir computing models, kernel-based learning for structured data, and applications. He currently serves as an Associate Editor of IEEE TNNLS.

Torna su ▲

When Data Science meets Process Intelligence: Process Mining

Dove / Data e Ora: 1BC45 (DM - Torre Archimede), il giorno 20/02/2019 alle ore 14:30

Relatore: Massimiliano de Leoni (Assistant Professor of Computer Science (RTD/b) at the Department of Mathematics, University of Padova)

Abstract: Since 2019, Massimiliano de Leoni is an Assistant Professor of Computer Science (RTD/b) at the Department of Mathematics, University of Padova. From September 2014 till December 2018, he was an Assistant Professor at the Technical University of Eindhoven (TU/e), the Netherlands. In 2009, he earned a PhD in Computer Engineering at La Sapienza University of Rome, Italy. He was a guest research fellow at Queensland University of Technology, Free University of Bolzano, Vienna University of Economics and Business, and University of Naples. In 2018, he was Visiting Professor at Università Politecnica delle Marche. His research interests are in the areas of Data Science, Process-aware Information Systems and Business Process Management, predominantly on Process Mining and Decision Support Systems.

CV Relatore: Processes are everywhere, when we enter a hospital or we send a package, when we are enrolled at university or we open a bank account. They are usually supported by information systems that record the processes’ executions in so-called event logs. In the new era of Big Data, these event logs are quickly becoming richer and richer: on the one hand, information systems are more pervasive (e.g., in home automation or logistics) and connected with external systems and, on the other hand, these event logs can be augmented with additional data that come from external “data factories”, including social media, geo-referenced physical objects (e.g. via RFID tags) and questionnaires. The large availability of process data is more than just a matter of volume and all the related challenges. Compared with traditional Business Intelligence, this is an opportunity to gain actionable insights to help organizations make better business decisions and become more effective and competitive. The field that crosses Business Intelligence, Process Management and Data Science is called Process Mining. It provides new means to discover, monitor and improve processes in a variety of application domains, based on the “real” facts recorded in the event logs rather than on the subjectivity of process stakeholders and owners. This talk will start with a brief introduction to Process Mining and continue illustrating some of the contribution of dr. de Leoni to the field. A number of real-life case studies will be discussed to, e.g., reduce costs, improve customer satisfaction or, also, diagnose unlawful deviations and detect their root-causes.

Torna su ▲

Answering questions in biology and medicine by making inferences on networks

Dove / Data e Ora: 2BC60 (DM - Torre Archimede), il giorno 11/02/2019 alle ore 10:00

Relatore: Prof. Alberto Paccanaro (Royal Holloway University of London)

Abstract: An important idea that has emerged recently is that a cell can be viewed as a set of complex networks of interacting bio-molecules and genetic disease is the result of abnormal interactions within these networks. In this talk, I'll present novel computational methods for answering questions in systems biology and medicine which can all be phrased in terms of inference and structure discovery in such large scale networks. These methods are based and extend recent developments in the areas of machine learning (particularly semi-supervised learning and matrix factorization), graph theory and network science. I’ll show how these computational techniques can provide effective solutions for: 1) quantifying similarity between heritable diseases at molecular level using exclusively disease phenotype information; 2) disease gene prediction; 3) drug side-effect prediction.

CV Relatore: Alberto Paccanaro is full Professor in Machine Learning and Computational Biology in the Department of Computer Science at Royal Holloway University of London where he is also Director of the Centre for Systems and Synthetic Biology. He completed his undergraduate studies in Computer Science at the University of Milan and received his PhD from the University of Toronto in 2002, specializing in machine learning under the supervision of Geoffrey Hinton. From 2002 to 2006 he was a postdoc in Mansoor Saqi’s lab at Queen Mary University of London and then in Mark Gerstein’s lab at Yale University. His research interests are in applying and developing machine learning and pattern recognition techniques for solving problems in molecular biology and medicine. His recent work has focused on the development of methods for analysis and inference in large scale biological networks. Labpage: www.paccanarolab.org

Torna su ▲

Interpretable preference learning: a game theoretic framework for large margin on-line feature and rule learning

Dove / Data e Ora: 2BC30 (DM - Torre Archimede), il giorno 29/11/2018 alle ore 15:00

Relatore: Mirko Polato (University of Padova)

Abstract: A large body of research is currently investigating on the connection between machine learning and game theory. In this work, game theory notions are injected into a preference learning framework. Specifically, a preference learning problem is seen as a two-players zero-sum game. An algorithm is proposed to incrementally include new useful features into the hypothesis. This can be particularly important when dealing with a very large number of potential features like, for instance, in relational learning and rule extraction. A game theoretical analysis is used to demonstrate the convergence of the algorithm. Furthermore, leveraging on the natural analogy between features and rules, the resulting models can be easily interpreted by humans. An extensive set of experiments on classification tasks shows the effectiveness of the proposed method in terms of interpretability and feature selection quality, with accuracy at the state-of-the-art.

CV Relatore: - - -

Torna su ▲

Using AI to Enhance Security

Dove / Data e Ora: Sala Riunioni 701, il giorno 26/11/2018 alle ore 13:00

Relatore: Prof. V.S. Subrahmanian (Dartmouth College)

Abstract: In this talk, Prof. Subrahmanian will describe 5 broad sets of efforts at the intersection of AI and security. In particular, he will first discuss work on identifying malicious actors in online platforms including bots, trolls, vandals, and sockpuppets. He will then discuss a set of efforts in which his group adapts machine learning techniques to significantly improve security of the Android platform and better understand Android behavior. Third, he will discuss work that uses game models to enhance security. Finally, he will discuss ongoing work on helping fight IP theft and some other works by his colleagues in the Institute for Security, Technology, and Society at Dartmouth.

CV Relatore: V.S. Subrahmanian is the Dartmouth College Distinguished Professor in Cybersecurity, Technology, and Society and Director of the Institute for Security, Technology, and Society at Dartmouth. He previously served as a Professor of Computer Science at the University of Maryland from 1989-2017 where he created and headed both the Lab for Computational Cultural Dynamics and the Center for Digital International Government. He also served for 6+ years as Director of the University of Maryland's Institute for Advanced Computer Studies (UMIACS). Prof. Subrahmanian is an expert on big data analytics including methods to analyze text/geospatial/relational/social network data, learn behavioral models from the data, forecast actions, and influence behaviors with applications to cybersecurity and counter-terrorism. He has written five books, edited ten, and published over 300 refereed articles. He is a Fellow of the American Association for the Advancement of Science and the Association for the Advancement of Artificial Intelligence and received numerous other honors and awards. His work has been featured in numerous outlets such as the Baltimore Sun, the Economist, Science, Nature, the Washington Post, American Public Media. He serves on the editorial boards of numerous journals including Science, the Board of Directors of the Development Gateway Foundation (set up by the World Bank), SentiMetrix, Inc., and on the Research Advisory Board of Tata Consultancy Services. He previously served on DARPA's Executive Advisory Council on Advanced Logistics and as an ad-hoc member of the US Air Force Science Advisory Board.

Torna su ▲

Online Conformance Checking

Dove / Data e Ora: 1BC50, il giorno 17/10/2018 alle ore 17:00

Relatore: Andrea Burattin (DTU (Denmark))

Abstract: Andrea Burattin is tenure-track assistant professor at DTU (Denmark). Before that he has been post doctoral researcher at the University of Innsbruck (Austria) and University of Padua. The IEEE Task Force on Process Mining awarded to his Ph.D. thesis the Best Process Mining Dissertation Award 2012-2013. He serves as organizer of Business Process Intelligence (BPI) workshop since 2015 and he is in the program committee of several conferences, such as BPM, ICPM, ICSOFT and Profes.

CV Relatore: The field of conformance checking, part of process mining, aims to quantify the extent to which the execution of a process, captured within recorded corresponding event data, conforms to a given reference model. Existing techniques assume a post-mortem scenario, i.e. they detect deviations based on complete executions of the process. Such assumption limits their applicability in an online setting. In such context, we aim to detect deviations online (i.e., in-vivo), in order to provide recovery possibilities before the execution of a process instance is completed. Online conformance checking carries scientific and computational challenges, since the processing has to take place immediately when events are emitted, possibly at very high speed. The benefits of these family of techniques, however, are impactful: to provide immediate knowledge on what is happening, thus detecting very early if the system is properly behaving and how its users interact with it. This talk will cover the basics of online process mining as well as technical details about online conformance checking. This talk is based on results of joint collaborations with Josep Carmona (UPC, Barcelona), Abel Armas-Cervantes (The University of Melbourne), Bas van Zelst (Fraunhofer Institute FIT, Aachen) and Boudewijn van Dongen (TU/Eindhoven).

Torna su ▲

CyberChallenge.IT 2018: A Summary

Dove / Data e Ora: 1AD100 , il giorno 11/07/2018 alle ore 15:00

Relatore: The CyberChallenge.IT team and local organizers

Abstract: More information about the event on this web page: http://spritz.math.unipd.it/seminars/2018/cyberchallenge.html

CV Relatore: - - -

Torna su ▲

Advances in deep generative models

Dove / Data e Ora: 2BC60, il giorno 20/06/2018 alle ore 12:00

Relatore: Nicola De Cao (University of Amsterdam)

Abstract: Deep Generative Models (DGMs) use unsupervised learning techniques to build a samplable function that matches any data distribution. The primary objective of a DGM is to learn the true data distribution from a set of data-points such that, subsequently, it is possible to generate new data points from a parametrized learned approximation of it. They have achieved enormous success in the past few years thanks to the use of deep learning techniques which allowed to learn effective distribution approximations and led to impressive results and applications in many fields. We will provide a brief introduction of two of the most commonly used approaches for designing DGMs namely: Variational Auto-Encoders (VAE) and Generative Adversarial Networks (GAN). We further introduce two recent advances in this field such as Hyperspherical VAE and MolGAN (an implicit model for drugs generation).

CV Relatore: Nicola De Cao is a future Ph.D. candidate at the Natural Language Processing Group at the University of Edinburgh starting next September. His work will be centered on developing models to address machine comprehension on unstructured documents. Currently, he is a master student in Artifical intelligence at University of Amsterdam (UvA). His thesis focuses on proposing and comparing explicit and implicit generative models for small molecular graphs [1]. Nicola recently worked with the Amsterdam Machine Learning Lab (AMLab) focusing on manifold learning for generative models proposing two novel models such as Hyperspherical [2] and Homeomorphic Variational Auto-Encoders. Previously, he was a bachelor student in Computer Science at University of Padua (2013-2016). [1] De Cao, Nicola, and Thomas Kipf. ''MolGAN: An implicit generative model for small molecular graphs.'' arXiv preprint arXiv:1805.11973 (2018). [2] Davidson, Tim R., et al. ''Hyperspherical Variational Auto-Encoders.'' arXiv preprint arXiv:1804.00891 (2018).

Torna su ▲

Autonomous Cyber Deception Systems

Dove / Data e Ora: 2BC60, il giorno 20/06/2018 alle ore 11:00

Relatore: Fabio De Gaspari (Sapienza University of Rome)

Abstract: Due to the increasingly sophisticated nature of cyber attacks, the effectiveness of classical defence systems and expert human intervention is heavily reduced. As a consequence, in recent years interest in active defence and intelligent, autonomous agents has grown considerably. This talk focuses on improvement of active cyber defence and its application to autonomous agents. Autonomous, cyber deception agents use active defence tools as sensors, and dynamically plan the deception strategy based on the world view provided by these sensors. Moreover, the agent also uses the active defence tools as actuators to effectively implement the plan devised. This architecture allows to actively hinder attackers progress, as well as to constantly update the defence plan based on the behaviour of the attacker.

CV Relatore: Fabio De Gaspari is a PhD student at the Sapienza University of Rome, Dipartimento di Informatica. His research areas are network security, Future Internet Architectures and intelligent active defence systems. Fabio obtained his MSc in computer science in 2015 from the University of Padova, Italy.

Torna su ▲

Binary Analysis Notes

Dove / Data e Ora: Torre Archimede 1A150, il giorno 21/06/2018 alle ore 14:00

Relatore: Mariano Graziano (Cisco)

Abstract: The ''Binary Analysis Notes'' seminar provides an overview of cybersecurity topics, while boosting interests for this interesting discipline. We will start with a short introduction about origins of hacking and will move to today's activities, such as Catch The Flag competitions and bug bounty rewards. We will deeply focus on binary analysis, anti-analysis and anti-debugging techniques. The seminar will be concluded with an exercise on real ELF binary unpacking. For registration and further info, please, visit here (spritz.math.unipd.it/events/2018/Cisco_Reversing/index.html)

CV Relatore: Mariano Graziano is a cybersecurity researcher. He obtained his PhD at Eurecom and now he works for Cisco in Talos group.

Torna su ▲

FinTech revolution: rischi e opportunità. Il punto di vista di Intesa Sanpaolo

Dove / Data e Ora: Aula 1BC50, il giorno 09/05/2018 alle ore 08:30

Relatore: Massimo Tessitore (Responsabile Direzione Multicanalità Integrata Intesa Sanpaolo)

Abstract: “Intelligenza artificiale”, “moneta elettronica” e “criptovalute” sono solo alcuni termini associabili al settore FinTech che sempre più di frequente popola le news dei vari canali informativi. Mai come oggi la tecnologia gioca un ruolo di primo piano nell’indicare i trend di sviluppo del mercato finanziario. La spinta all’innovazione nel mondo bancario è diventata ormai un imperativo e Intesa Sanpaolo, la più importante banca italiana, ha da tempo avviato questo processo abbracciando la trasformazione digitale in tutte le sue declinazioni come fattore chiave abilitante e facendolo diventare un importante pilastro del loro modello di servizio e del loro modello di business.

CV Relatore: - - -

Torna su ▲

Underwater Communications and Networking

Dove / Data e Ora: SRVII, il giorno 11/04/2018 alle ore 13:30

Relatore: Roberto Petroccia (NATO STO Centre for Maritime Research and Experimentation (CMRE))

Abstract: Underwater networks are an enabling technology in a wide spectrum of different collaborative scenarios that find application to science, security, and industry. In the underwater domain, acoustics is the main technology used so far for communications, since radio and optical signals are greatly attenuated. Nonetheless acoustic communications in water are characterised by many challenges that are specific for the considered underwater environment: Long propagation delays, low bandwidth, sound speed variability, slow power signal attenuation, and many other environmental impairments. The wireless nature of the communications makes underwater acoustic networks vulnerable to various malicious attacks, yet, limited consideration has been given to the security of such networks. This talk presents the main challenges of underwater communication and networking, addressing also the current state of the art of security solutions in the underwater domain. Future steps towards software-defined strategies and collaborative approaches are introduced.

CV Relatore: Roberto Petroccia received a Laurea Degree with the highest honours (2006) and a Ph.D. (2010), both in Computer Science, from Rome University “La Sapienza,” Italy, where he has been also affiliated as research staff until 2015. Since 2015, Dr. Petroccia is a Research Scientist at the NATO STO Centre for Maritime Research and Experimentation (CMRE). His research interests include wireless sensor networks design, underwater communications and networking along with autonomy of robots and standardisation, where he has contributed over three-dozens papers published in leading venues. In the last five years Dr. Petroccia has participated to over twenty experimental campaigns at sea where innovative underwater solutions he developed have been extensively tested. He has been actively collaborating with several acoustic modem and underwater vehicle manufacturing companies and research labs to design novel technologies supporting cooperative underwater acoustic networks. Dr. Petroccia is an invited lecturer of the Masters in Ocean Engineering offered by the University of Pisa (ITA) and of the Submariner Specialisation Course by the Italian Navy Academy of Livorno (ITA). He has supervised the work of several master thesis and Ph.D. students, he is also a Senior member of IEEE.

Torna su ▲

Seminario tenuto da Luca Moroni - Via Virtuosa

Dove / Data e Ora: 1A150 in Torre Archimede, il giorno 23/02/2018 alle ore 09:30

Relatore: Luca Moroni (Via Virtuosa)

Abstract: https://www.informazione.it/c/22A912CE-1B67-4105-81E3-07D01DA169E2/Via-Virtuosa-La-prima-startup-innovativa-che-seleziona-i-fornitori-di-servizi-in-ambito-Cyber-Security

CV Relatore: - - -

Torna su ▲

Digital Forensics: Overview and Research Issues

Dove / Data e Ora: 1C150, Torre Archimede, il giorno 22/02/2018 alle ore 10:00

Relatore: Professor Antonio Barili (Universita' di Pavia)

Abstract: On August, 2001 the first Digital Forensic Research Workshop (DFRWS) was held in Utica, New York. The workshop was attended by over 50 researchers, computer forensic examiners, and analysts. Its major goal was to establish a research community that would apply the scientific method in finding focused near-term solutions driven by practitioner requirements and addressing longer term needs. More than 15 years later, despite the impressive advances in the field, the number of technical and scientific issues has grown bigger and LEAs' forensic labs suffer from an unprecedented technical resource shortage and increasing backlogs. The seminar will provide an overview and 'state of the art' of the field of Digital Forensics and address a few of the current research issues, namely those arising from the huge amount of data to be processed and from 'semantic' text mining.

CV Relatore: Antonio Barili is a Researcher and Principal Investigator at the Digital Forensic Lab of the University of Pavia. As a forensic expert, he handled more than a hundred cases in the last ten years, ranging from computer frauds to murder.

Torna su ▲

Deep Detection Architecture for Security in Industrial Control Systems

Dove / Data e Ora: Aula 2AB45 in Torre Archimede, il giorno 02/02/2018 alle ore 14:30

Relatore: Ing. Giuseppe Bernieri (Deloitte Cyber Security Consultant and PhD Candidate Department of Engineering University ''Roma Tre'', Italy)

Abstract: In recent years, the evolution of information and communication technology joined the industrial control systems development, leading to new significant enhancements. Even though the benefits are noticeable, new security challenges concerning industrial facilities arise: typical vulnerabilities of the cyber domain emerged in industrial control systems. In this context, the classic cyber-security tools are ineffective, since industrial control systems are designed to operate in standalone or isolated configurations and are characterized by hard real-time constraints. As a consequence, over the last decade, critical infrastructures have experienced a large number of cyber-attacks. Despite the impact of such alarms, the paramount importance of the information traveling in the control system networks and its protection is still underestimated. In this seminar, the protection of industrial control systems for critical infrastructures will be addressed. The Deep Detection Architecture is a modular framework conceived to provide security in the Control Zone and can be implemented in a distributed fashion. It is composed by three main systems, the Deep Detection System (DDS), the Mimepot, and the Decision Support System (DSS). The proposed architecture fills the gap between computer science and control theoretic approaches. To validate the proposed architecture, several tools have been developed and will be presented.

CV Relatore: Giuseppe Bernieri is actually working for Deloitte as Cyber Security Consultant. He completed his PhD in Computer Science and Automation at the Dept. of Engineering of the University of Roma Tre. His interests focus on cyber security applied to industrial control systems and critical infrastructures, the development, and implementation of ad-hoc solutions for the detection of cyber-physical threats affecting SCADA systems. He collaborates with the Network, Information and Computer Security (NICS) Lab of Malaga (Spain) and with the iTrust research centre located at the Singapore University of Technology and Design (SUTD).

Torna su ▲

Logic Tensor Networks for Semantic Image Interpretation

Dove / Data e Ora: sala riunioni DEI/G, piano 3 DEI/G Dipartimento di Ingegneria dell'Informazione (Via Gradenigo 6a), il giorno 07/11/2017 alle ore 16:30

Relatore: Dr. Ivan Donadello (Fondazione Bruno Kessler, University of Trento)

Abstract: Semantic Image Interpretation (SII) is the task of extracting structured semantic descriptions from images. This task is suitable for Logic Tensor Networks (LTNs), a new Statistical Relational Learning framework which integrates neural networks with fuzzy logic. LTNs allows learning and reasoning with data and logical constraints. In this seminar, we show LTNs and its evaluation on the classification of image objects and part-of relations between them. The use of logical constraints improves the performance of purely data-driven approaches. Moreover, the logical constraints add robustness when the training labels are affected by errors. http://www.dei.unipd.it/colloquia

CV Relatore: - - -

Torna su ▲

Security Issues of Mobile and Smart Wearable Devices

Dove / Data e Ora: SRVII, il giorno 22/11/2017 alle ore 16:30

Relatore: Hossein Fereidooni (Università degli Studi di Padova / TU Darmstadt)

Abstract: Mobile and smart devices (ranging from popular smartphones and tablets to wearable fitness trackers equipped with sensing, computing and networking capabilities) have proliferated lately and redefined the way users carry out their day-to-day activities. These devices bring immense benefits to society and boast improved quality of life for users. As mobile and smart technologies become increasingly ubiquitous, the security of these devices becomes more urgent, and users should take precautions to keep their personal information secure. Privacy has also been called into question as so many of mobile and smart devices collect, process huge quantities of data, and store them on the cloud as a matter of fact. Ensuring confidentiality, integrity, and authenticity of the information is a cybersecurity challenge with no easy solution. This talk deals with the security problems of mobile and smart devices, providing specific methods for improving current security solutions. In the first part of this talk, we study methods and techniques to assist security analysts to tackle mobile malware and automate the identification of malicious applications. First, we introduce a Secure Message Delivery (SMD) protocol for Device-to-Device (D2D) networks, with primary objective of choosing the most secure path to deliver a message from a sender to a destination in a multi-hop D2D network. Second, we illustrate a survey to investigate concrete and relevant questions concerning Android code obfuscation and protection techniques, where the purpose is to review code obfuscation and code protection practices. Finally, we propose a Machine Learning-based detection framework to hunt malicious Android apps by introducing a system to detect and classify newly-discovered malware through analyzing applications. The second part of the talk conducts an in-depth security analysis of the most popular wearable fitness trackers on the market. First, we analyze the primitives governing the communication between fitness tracker and cloud-based services. In addition, we investigate communication requirements in this setting such as: (i) Data Confidentiality, (ii) Data Integrity, and (iii) Data Authenticity. Second, we show real-world demos on how modern wearable devices are vulnerable to false data injection attacks. Also, we document successful injection of falsified data to cloud-based services that appears legitimate to the cloud to obtain personal benefits. Third, we circumvent End-to-End protocol encryption implemented in the most advanced and secure fitness trackers (e.g., Fitbit, as the market leader) through Hardware-based reverse engineering. Last but not least, we provide guidelines for avoiding similar vulnerabilities in future system designs.

CV Relatore: Hossein Fereidooni received his BS.c. and MS.c. degrees in Biomedical and Electrical Engineering, from Poly technique Tehran, Iran, in 2007 and 2009, respectively. In 2014, He joined to SPRITZ Security and Privacy Research Group, University of Padua, as a PhD student under supervision of Prof. Mauro Conti. He currently is working as a Scientific Researcher for The Cybersecurity Research Center (CYSEC), an internationally known and one of the largest in Europe with more than 280 researchers only devoted to security and privacy research, at Technical University Darmstadt in Germany.

Torna su ▲

Process Mining: Data Science in Action

Dove / Data e Ora: Aula 1A150 Torre Archimede, il giorno 09/10/2017 alle ore 16:30

Relatore: Wil van der Aalst (Technische Universiteit Eindhoven - TU/e)

Abstract: Process mining provides new ways to utilize the abundance of event data in our society. This emerging scientific discipline can be viewed as a bridge between data science and process science: It is both data-driven and process-centric. Process mining provides a novel set of techniques to discover the real processes. These discovery techniques return process models that are either formal (precisely describing the possible behaviors) or informal (merely a “picture” not allowing for any form of formal reasoning). Formal models are able to classify traces (i.e., sequences of events) as fitting or non-fitting. Most process mining approaches described in the literature produce such models. This is in stark contrast with the over 25 available commercial process mining tools that only discover informal process models that remain deliberately vague on the precise set of possible traces. There are two main reasons why vendors resort to such models: scalability and simplicity. In this talk, prof. Van der Aalst will propose to combine the best of both worlds: discovering hybrid process models that have formal and informal elements. The discovered models allow for formal reasoning, but also reveal information that cannot be captured in mainstream formal models. A novel discovery algorithm returning hybrid Petri nets has been implemented in ProM and will serve as an example for the next wave of commercial process mining tools. Prof. Van der Aalst will also elaborate on his collaboration with industry. His research group at TU/e applied process mining in over 150 organizations, developed the open-source tool ProM, and influenced the 20+ commercial process mining tools available today.

CV Relatore: Prof.dr.ir. Wil van der Aalst is a distinguished university professor at the Technische Universiteit Eindhoven (TU/e) where he is also the scientific director of the Data Science Center Eindhoven (DSC/e). Since 2003 he holds a part-time position at Queensland University of Technology (QUT). Currently, he is also a visiting researcher at Fondazione Bruno Kessler (FBK) in Trento and a member of the Board of Governors of Tilburg University. His personal research interests include process mining, Petri nets, business process management, workflow management, process modeling, and process analysis. Wil van der Aalst has published over 200 journal papers, 20 books (as author or editor), 450 refereed conference/workshop publications, and 65 book chapters. Many of his papers are highly cited (he one of the most cited computer scientists in the world; according to Google Scholar, he has an H-index of 135 and has been cited 80,000 times) and his ideas have influenced researchers, software developers, and standardization committees working on process support. Next to serving on the editorial boards of over 10 scientific journals he is also playing an advisory role for several companies, including Fluxicon, Celonis, and ProcessGold. Van der Aalst received honorary degrees from the Moscow Higher School of Economics (Prof. h.c.), Tsinghua University, and Hasselt University (Dr. h.c.). He is also an elected member of the Royal Netherlands Academy of Arts and Sciences, the Royal Holland Society of Sciences and Humanities, and the Academy of Europe. Recently, he was awarded with a Humboldt Professorship, Germany’s most valuable research award (five million euros), and will move to RWTH Aachen University at the beginning of 2018.

Torna su ▲

How to Bypass Memory Protection: Evolution of Return-Oriented Programming and Rowhammer Attacks

Dove / Data e Ora: 1BC45, il giorno 05/10/2017 alle ore 08:30

Relatore: Lucas Davi (University of Duisburg-Essen)

Abstract: Memory corruption attacks such as return-oriented programming are a powerful exploitation technique to compromise software on a wide range of architectures. These attacks generate malicious computation based on existing code (so-called gadgets) residing in linked libraries. Both academia and industry have recently proposed defense techniques to mitigate return-oriented programming attacks. However, a continuous arms race has evolved between attacks and defenses. In this talk, we will elaborate on the evolution of memory corruption attacks. In particular, we explore prominent defense techniques such as control-flow integrity (CFI) enforcement, code randomization, and remote attestation. In addition, we investigate memory corruption attacks that exploit hardware faults to induce dangerous bit flips in memory. These attacks undermine memory access control mechanisms without requiring any software vulnerability. Specifically, we elaborate on the evolution of Rowhammer attacks and defenses.

CV Relatore: Lucas Davi is assistant professor for secure software systems at University of Duisburg-Essen, Germany. He received his PhD from TU Darmstadt in computer science. His research focus includes aspects of system security, software security, and trusted computing, especially software exploitation techniques and defenses. He received best paper awards at DAC, ACM ASIACCS, and IEEE Security and Privacy. His PhD thesis on code-reuse attacks and defenses has been awarded with the ACM SIGSAC Dissertation Award 2016.

Torna su ▲

IoT - ''next digital revolution'' and 5G networks

Dove / Data e Ora: SRVII, il giorno 20/10/2017 alle ore 10:00

Relatore: Dr. Haitham S. Cruickshank (University of Surrey)

Abstract: IoT is one of the major application areas in 5G Networks. As such, there is a growing belief that the Internet of Things (IoT) represents the start of the next digital revolution, where everyday objects are connected to a network in order to share their data. By 2020, the global value of the IoT sector is estimated to exceed $400 billion per year. The emerging IoT applications rely increasingly on the capture and processing of personal data. Two examples are Connected and Autonomous Vehicles (CAV) and Intelligent Transport Systems (ITS), as a major transport application for IoT sensors, where the question of security and consent in the use of personal data should be resolved. The lecture presents research challenges in for privacy and security, where decentralised and distributed solutions might help for such large scale CAV/ITS systems together with low impact on the underlying 5G network performance.

CV Relatore: http://www.ee.surrey.ac.uk/Personal/H.Cruickshank

Torna su ▲

Randomization Can't Stop BPF JIT Spray

Dove / Data e Ora: SRVII, il giorno 21/09/2017 alle ore 15:00

Relatore: Filippo Bonazzi (Aalto University)

Abstract: Berkeley Packet Filter (BPF) is a mechanism introduced in the Linux kernel as a way to perform fast line-speed filtering of network packets. To achieve sufficient filtering speeds, a Just-In-Time compiler was added to BPF, effectively transforming it into a general-purpose mechanism to safely support interpreted code injected into a Linux kernel. The JIT compiler was attacked back in 2012 by Keegan McCallister, who used it to inject an exploit payload. Upstream Linux mitigations adopted in response were limited to randomization. This talk will present a modified proof of concept that demonstrates the possibility of a successful BPF JIT spray attack on the 4.4 upstream Linux kernel, and discuss the appropriate mitigations which have since been merged in the 4.7 upstream Linux kernel.

CV Relatore: Filippo Bonazzi is a researcher interested in Linux kernel security, platform security, cryptography, malware analysis and network security. His past research includes work on SEAndroid and V2X privacy, and he is currently working on embedded system security and privacy. Filippo obtained his Master's Degree in Computer Engineering from Politecnico di Torino, Italy, and spent two years as a researcher in the Secure Systems Group at Aalto University, Helsinki, Finland.

Torna su ▲

Securing Databases from Probabilistic Inference

Dove / Data e Ora: SRVII, il giorno 12/09/2017 alle ore 11:00

Relatore: Marco Guarnieri (ETH Zurich)

Abstract: Databases can leak confidential information when users combine query results with probabilistic data dependencies and prior knowledge. Current research efforts offer mechanisms that either handle a limited class of dependencies or lack tractable enforcement algorithms necessary for scaling. We propose a foundation for Database Inference Control based on PROBLOG, a probabilistic logic programming language. We leverage this foundation to develop ANGERONA, a provably secure enforcement mechanism that prevents information leakage in the presence of probabilistic dependencies. We then provide a tractable inference algorithm for a practically relevant fragment of PROBLOG. We empirically evaluate ANGERONA's performance showing that it scales to relevant problems of interest.

CV Relatore: Marco Guarnieri is a 5th year PhD student in the Institute of Information Security at ETH Zurich under the supervision of Prof. David Basin. In his research, he combines techniques and concepts from various domains, such as database theory and programming languages, to build provably secure systems, with a focus on database security.

Torna su ▲

Bug Bounty Platforms: Empirical Analysis and Economic Challenges

Dove / Data e Ora: SRVII, il giorno 11/09/2017 alle ore 11:00

Relatore: Jens Grossklags (TU München)

Abstract: Despite significant progress in software-engineering practices, software utilized for web and mobile computing remains insecure. At the same time, the consumer and business information handled by these programs is growing in its richness and monetization potential, which triggers significant privacy and security concerns. In response to these challenges, companies are increasingly harvesting the potential of external (ethical) security researchers through bug bounty programs to crowdsource efforts to find and ameliorate security vulnerabilities. More recently, several commercial bug bounty platforms have emerged (e.g., HackerOne, BugCrowd, Cobalt, Wooyun) which successfully facilitate the process of building and maintaining bug bounty programs for organizations. To cite just one success story, on HackerOne, more than 40,000 security vulnerabilities have been reported and fixed for hundreds of organizations. In this talk, I will discuss our research over the last three years which systematically studies these platforms. In particular, I will present empirical results demonstrating the growing popularity and practical contributions of two of these platforms, HackerOne and Wooyun. Unfortunately, the data also reveals a number of economic challenges which may limit the success of these platforms in the future. To respond to these hurdles, I will discuss different economic policies to improve their efficiency. I will close with a conversation about pressing policy considerations. The talk is based on joint work with Mingyi Zhao, Aron Laszka, and Thomas Maillart.

CV Relatore: Jens Grossklags is Professor of Cyber Trust in the Department of Informatics at the Technical University of Munich. Previously, he directed the Security, Privacy and Information Economics Lab, and served as the Haile Family Early Career Professor at Penn State. He was a Postdoctoral Research Associate at the Center for Information Technology Policy at Princeton University, and received his Ph.D. from UC Berkeley. He studies security and privacy challenges from the economic and behavioral perspectives with a variety of methodologies.

Torna su ▲

Eavesdropping LTE communications: from network optimization to security concerns

Dove / Data e Ora: 1BC45, il giorno 04/09/2017 alle ore 11:00

Relatore: Nicola Bui (Northeastern University, Boston)

Abstract: My recent studies focused on resource allocation optimization for LTE network. In such a framework, I mainly adopted a set of linear programming formulations to investigate different aspects, such as resource savings, quality-of-service maximization and admission control. All the proposed concepts have been evaluated over real LTE traffic recorded in four locations in Madrid over a month. During the talk, I am going to give a brief overview of the results obtained and the measurement campaign. Then, I will illustrate a few directions I am exploring at NEU: in particular, those regarding security and privacy vulnerabilities of current and future networks.

CV Relatore: Nicola Bui received his Master’s of Science and Ph.D. degrees in Telematics from Carlos III University of Madrid, Spain in 2014 and 2017, respectively. He is a Senior Research Scientist at Northeastern University in Boston. His current research focuses on 5G mobile networks and related security and privacy aspects. Prior to joining Northeastern, Bui worked as a Research Engineer at IMDEA Networks Institute in Madrid, Spain. In 2015 Bui was a visiting researcher at Nokia Bell Labs in Stuttgart. From 2006 to 2013, he was the CEO at Patavina Technologies, a spin-off of the University of Padova in Italy, developing embedded systems. At the same time, he collaborated with Consorzio Ferrara in Ricerche, Italy and with the Department of Information Engineering at the University of Padova in Italy. During this time, he contributed to many European and Italian projects such as e-SENSE, SENSEI, IoT-A, WISEWAI and SWAP. Bui authored more than 50 conference and journal papers.

Torna su ▲

Towards High-Interaction Virtual ICS Honeypots-in-a-Box

Dove / Data e Ora: SRVII, il giorno 22/08/2017 alle ore 14:00

Relatore: Daniele Antonioli (Singapore University of Technology and Design)

Abstract: In this work, we address the problem of designing and implementing honeypots for Industrial Control Systems (ICS). Honeypots are vulnerable systems that are set up with the intent to be probed and compromised by attackers. Analysis of those attacks then allows the defender to learn about novel attacks and general strategy of the attacker. Honeypots for ICS systems need to satisfy both traditional IT requirements, such as cost and maintainability, and more specific ICS requirements, such as time and determinism. We propose the design of a virtual, high-interaction and server- based ICS honeypot to satisfy the requirements, and the deployment of a realistic, cost-effective, and maintainable ICS honeypot. An attacker model is introduced to complete the problem statement and requirements. Based on our design and the MiniCPS framework, we implemented a honeypot mimicking a water treatment testbed. To the best of our knowledge, the presented honeypot implementation is the first academic work targeting Ethernet/IP based ICS honeypots, the first ICS virtual honeypot that is high-interactive without the use of full virtualization technologies (such as a network of virtual machines), and the first ICS honeypot that can be managed with a Software-Defined Network (SDN) controller.

CV Relatore: Daniele holds a BS and MS in Electronics and Telecommunications Engineering from the University of Bologna (UniBO, Italy). He started working on hardware security while preparing his master thesis about random number generation at the University of Massachussets (UMass, Amherst) advised by W. Burleson and R. Rovatti. Currently he is doing a PhD in Computer Science at the Singapore University of Design and Technology (SUTD, Singapore) advised by N.O. Tippenhauer. He is working on industrial control systems security. His main research interests are: applied security for cyber-physical systems, hardware security for embedded systems and physical layer security of wireless systems.

Torna su ▲

Location-privacy in the mobile era: challenges and solutions

Dove / Data e Ora: aula 1BC45, il giorno 27/06/2017 alle ore 11:00

Relatore: Luca Calderoni (Universita di Bologna)

Abstract: The evolution of mobile devices and the diffusion of location-based services (LBS), establish a cornerstone in the digital era. Traditional information (as names, addresses and phone numbers) running across the internet and through a number of other services are now coupled with positional data. With such an impressive amount of information, service providers are able to determine precise location estimates and to know almost everything concerning our opinions, religious and political preferences and private life in general. Anonymous and secure location-based services are thus an important challenge we need to face. Spatial Bloom Filters, a probabilistic data structure designed with location-privacy in mind, represent a viable solution for this issue.

CV Relatore: After a period spent at the Delft University of Technology, The Netherlands, he received a Ph.D. degree in computer science from the University of Bologna, Italy. He is currently a Post-doctoral Researcher with the Smart City Laboratory of the University of Bologna, in Cesena, Italy. His research activity focuses on privacy and security in digital systems and smart cities. Specifically, he published several research papers concerning location privacy, border controls, secure and privacy-preserving tracking and monitoring technologies, location-aware applications and urban ICT infrastructures.

Torna su ▲

You Surf So Strange Today: Anomaly Detection in Web Services via HMM and CTMC

Dove / Data e Ora: aula 1AD100, il giorno 05/05/2017 alle ore 16:00

Relatore: Maddalena Favaretto (dipartimento di matematica)

Abstract: In recent years, with the increasing number of attacks against user privacy in web services, researchers put a significant effort on realizing more and more sophisticated Intrusion Detection Systems in order to identify potentially malicious activities. Among such systems, Anomaly Detection Systems rely on a baseline given by a normal behavior and consider every deviation from such behavior as an intrusion. In this paper, we propose a novel Anomaly Detection System to detect intrusions in users’ private areas in on-line web services. Such services usually record logs of user activity from different points: access, actions in a session and system responses. We design an ad-hoc mathematical model for each of these logs to build a profile for a normal behavior. In particular, we model users’ accesses through a Hidden Markov Model (HMM) and Users’ activity with a Continuous Time Markov Chain (CTMC). We propose a novel Anomaly Detection System algorithm that takes into consideration the deviation from the above Markov Processes. Finally, we evaluate our proposal with a thorough set of experiments, which results confirm the feasibility and effectiveness of our solution.

CV Relatore: - - -

Torna su ▲

Walking backwards: x86 binary reversing 101

Dove / Data e Ora: aula 1 AD100, il giorno 05/05/2017 alle ore 14:30

Relatore: Andrea Biondo (studente di informatica del dipartimento)

Abstract: From undocumented interfaces to hidden bugs, closed-source software can hold plenty of surprises for whoever dares to look under the hood. Whether you need to work with legacy code, want to audit for security issues or are just plain curious, being able to dissect a piece of software and understand how it works is an invaluable skill to add to your toolbox. Despite the popularity of interpreted languages, most modern software is shipped in compiled, binary form. As such, a good understanding of binary reverse engineering is required to pry into it when the source code is not available. In this talk I will introduce you to the basics of binary reversing on 32- and 64-bit x86 platforms. I will begin with a refresher on the architecture and basic assembly. I will then walk you through recognizing higher level constructs and making sense of the low level code. Finally, I will introduce Hex-Rays IDA, industry standard and my tool of choice, and demonstrate real-world binary analysis and reversing with it.

CV Relatore: - - -

Torna su ▲

Sharing Knowledge for Large Scale Visual Recognition

Dove / Data e Ora: aula 2BC30 Torre Archimede, il giorno 30/03/2017 alle ore 12:00

Relatore: Lamberto Ballan (Dipartimento di Matematica Universita' di Padova)

Abstract: This talk overviews my research activities in computer vision, pattern recognition and multimedia for understanding big visual data. In particular, I will focus on two models for ''sharing'' prior and contextual knowledge for solving large scale visual recognition problems. In the first part of the talk, I'll show that images that are very difficult to recognize on their own may become more clear in the context of a neighborhood of related images with similar social-network metadata. Our model uses image metadata non-parametrically to generate neighborhoods of related images, then uses a deep neural network to blend visual information from the image and its neighbors. In the second part of the talk, I'll present our recent work on knowledge transfer for scene-specific motion prediction. When given a single frame of a video, humans can not only interpret the content of the scene, but also they are able to forecast the near future. This ability is driven by their rich prior knowledge about the visual world, both in terms of the dynamics of moving agents, as well as the semantic of the scene. We exploit the interplay between these two key elements to predict scene-specific motion patterns on a novel large dataset collected from UAV.

CV Relatore: - - -

Torna su ▲

Variational Markov Logic

Dove / Data e Ora: Aula 1C150 Torre Archimede, il giorno 17/03/2017 alle ore 11:00

Relatore: Radim Nedbal (assegnista di ricerca presso il Dipartimento di Matematica / FBK Trento)

Abstract: A novel framework for statistical relational learning. We name it variational Markov logic to stress the similarity and also the main difference from Markov logic. In both, a first-order logic theory is mapped to a probability distribution function (PDF) over interpretations of the theory. Unlike Markov logic, which yields maximum a posteriori or maximum-likelihood approximation of the PDF using sampling methods, variational Markov logic yields the Bayesian posterior PDF inferred by variational techniques in a purely symbolic way, avoiding numeric calculations. This is crucial for efficiency, and it is due to the representation of the domain model as a variational Markov logic network. The goal of the presentation is to give a high level insight into the problem in terms of main concepts of SRL: Language, Model, Representation, Interpretation, Query, Algorithms. In particular, I'd like to spend more time with Algorithms based on variational techniques and point out opportunities for future improvement.

CV Relatore: - - -

Torna su ▲

Learning and reasoning with knowledge and data: a review of my favourite approaches

Dove / Data e Ora: aula 1BC45, Torre Archimede, il giorno 15/12/2016 alle ore 15:30

Relatore: Luciano Serafini (Fondazione Bruno Kessler, Trento Italy)

Abstract: Hybrid domains are domains where objects are organised in a structure (e.g., a labelled graph) and some of the components of such a structure is associated to a set of numerical attributes or features (e.g., the vertex of a graph are associated with numeric features, and the arcs are associated to weights). In these domains, structural properties and numerical properties are tightly connected and they cannot be managed separately. On the one hand, logical approaches provide excellent tools to describe some known structural properties of a domain and to automatically infer via deductive reasoning new true properties about the structures that logically follows from them. On the other hand, machine learning techniques, such as regression, kernels, support vector machines, neural networks and graphical models, are quite useful and flexible methodologies to infer, via inductive reasoning (aka learning), new numerical and structural properties from the numerical attributes/features associated to the domain. Since it's beginning Artificial Intelligence dream has been to find a satisfactory integration of these two forms of inference. Along the years many proposals have been done, but they never get to the state of being mature enough. It's only in the recent years, that researchers were looking to suitably combine Logical reasoning and machine learning in hybrid domains, and they developed a number of frameworks which seem to be promising for the solution of such a key AI challenge. In this talk I'll revise some of them. In particular, I'll give an overview of the main principles, which is at the base of all the modern approaches, and I'll briefly present some of the emerging approaches.

CV Relatore: - - -

Torna su ▲

Temporal logic and formal verification for cyber-physical systems

Dove / Data e Ora: aula 2AB 40 Torre Archimede, il giorno 17/11/2016 alle ore 11:30

Relatore: Davide Bresolin (Universita' di Padova)

Abstract: Cyber-Physical systems are characterized by the tight integration of cyber aspects (computing) with physical ones (e.g., mechanical, electrical, and chemical processes). Examples of such systems are autonomous robotic systems, multi-agent and embedded systems, automotive, aerospace and medical systems. They need to operate under strong safety, performance, reliability and timing constraints. They are characterized by a mixed discrete and continuous behaviour that cannot be characterized faithfully using either discrete or continuous models only, and thus they need the development of new formalisms and new algorithmic techniques to be formally verified. This talks overviews my research activity in developing temporal logics and reachability analysis techniques for the verification of cyber-physical systems.

CV Relatore: - - -

Torna su ▲

Learning 3D shape correspondence with Anisotropic CNNs

Dove / Data e Ora: Aula 1BC45 Torre Archimede, il giorno 19/09/2016 alle ore 14:30

Relatore: Emanuele Rololà (USI Lugano, CH)

Abstract: Establishing correspondence between shapes is a fundamental problem in geometry processing, arising in a wide variety of applications. The problem is especially difficult in the setting of non-isometric deformations, as well as in the presence of topological noise and missing parts, mainly due to the limited capability to model such deformations axiomatically. Several recent works showed that invariance to complex shape transformations can be learned from examples. In this paper, we introduce an intrinsic convolutional neural network architecture based on anisotropic diffusion kernels, which we term Anisotropic Convolutional Neural Network (ACNN). In our construction, we generalize convolutions to non-Euclidean domains by constructing a set of oriented anisotropic diffusion kernels, creating in this way a local intrinsic polar representation of the data (-patch-), which is then correlated with a filter. Several cascades of such filters, linear, and non-linear operators are stacked to form a deep neural network whose parameters are learned by minimizing a task-specific cost. We use ACNNs to effectively learn intrinsic dense correspondences between deformable shapes in very challenging settings, achieving state-of-the-art results on some of the most difficult recent correspondence benchmarks

CV Relatore: Emanuele Rodolà is a post-doctoral researcher at Università della Svizzera Italiana (USI Lugano) since February 2016, where he works in the group led by Prof. Michael Bronstein. Before that, he was an Alexander von Humboldt Fellow in Prof. Daniel Cremers' Computer Vision lab at TU Munich (2013-2016) and a JSPS Research Fellow at The University of Tokyo (Intelligent Systems and Informatics Lab, 2013). He received his PhD in Computer Science under the supervision of Prof. Andrea Torsello at Università Ca' Foscari Venezia (2012), and graduated in Computer System Engineering at the University of Rome ''Tor Vergata'' (2008). During his doctoral studies he spent a visiting research period at Tel Aviv University under the supervision of Prof. Alex Bronstein. He received a number of awards, including the Best Student Paper Award at 3DPVT 2010, the Best Paper Award at VMV 2015, and the Best Paper Award at SGP 2016. He has been serving in the program committees of the top rated conferences in computer vision (CVPR, ICCV, ECCV, ACCV, etc.), served as Area Chair at 3DV 2016, founded and chaired the first ECCV workshop on Geometry Meets Deep Learning (GMDL 2016), organized two SHREC 2016 contests, and was recognized as Outstanding Reviewer at CVPR (2013, 2015, 2016), ICCV (2015), and ECCV (2014). He gave tutorials and short courses in multiple occasions at EUROGRAPHICS, ECCV, and SIGGRAPH Asia. His work on 3D reconstruction was featured by the national Italian television (RAI - Cose dell'altro Geo) in 2012.

Torna su ▲

Static analysis for security at the Facebook scale

Dove / Data e Ora: aula 2AB45 Torre Archimede, il giorno 21/07/2016 alle ore 11:30

Relatore: Francesco Logozzo (Facebook inc., Seattle, USA)

Abstract: The scale and continuous growth of commercial code bases are the greatest challenges for adoption of automated analysis tools in Industry. Alas, scale is largely ignored by academic research. We developped a new static analysis tool for security to scale to Facebook scale. It relies on abstract interpretation to focus on the properties that really matter to security engineers and provides fine control on the cost/precision ratio. It was designed from day one for “real world” security and privacy problems at scale. Facebook codebase is huge, and we can analyze it, from scratch in 10 minutes. This talk will give attendees a peek at some of the secret sauce we use to achieve such amazing performance and precision

CV Relatore: - - -

Torna su ▲

Invariance proof methods for weakly consistent parallelism

Dove / Data e Ora: Aula 2AB45 Torre Archimede, il giorno 21/07/2016 alle ore 10:00

Relatore: Patrick Cousot (New York University, USA)

Abstract: We design an invariance proof method for concurrent programs parameterised by a weak consistency model. This generalises Lamport/Owicki-Gries method for sequential consistency. We use the cat language to write specifications of consistency models as well as concurrent program specifications.

CV Relatore: - - -

Torna su ▲

Cyber warfare Trends and Future

Dove / Data e Ora: Meeting Room VII Floor, il giorno 01/06/2016 alle ore 13:00

Relatore: Dr. Ali Dehghantanha (University of Salford UK)

Abstract: When Sir Timothy John Tim Berners-Lee envisioned the Internet, no one could probably though as how computers and computing clouds can be used as war instruments! However, with increasing reliance of governments and enterprises on computing devices this probability has become reality. Through out history of classical wars, civilian commercial entities have not been the primary targets of warfare! In those days the only means of warfare was kinetic warfare, using spears, swords, ballistic weapons, explosives, and so on. However, nowadays we are observing the appearance of doctrine and dedicated cyber warfare programs all around the world! Countries are begun to include cyber warfare in their military doctrine as well as their college curriculums on both offensive and defensive applications. They are developing strategies and tools to conduct information attacks and train developers as software soldiers while their main targets are usually normal civilians at least as jumping points! This talk lightens the current trends and mechanisms in cyber warfare with specific focus on countries that are actively involved and their attack techniques. This talk looks to the future directions in running such cyber attacks and necessity to educate users. Finally it suggests some solutions to make more peaceful cyber world.

CV Relatore: - - -

Torna su ▲

Finding the Needle in the Internet of Everythings Haystack

Dove / Data e Ora: Meeting Room VII Floor, il giorno 31/05/2016 alle ore 13:00

Relatore: Dr. Ali Dehghantanha (University of Salford UK)

Abstract: With the fast integration of smart things we are swiftly moving towards a pervasive, intelligent and integrated environment where smart sensors are collecting large amount of potentially private data. Internet of everything (IoT) would soon pervade all aspects of our life from managing our home temperature to thinking cars and smart management of the cities. So it won’t take long to see people suing others for misusing their smart things, thinking cars that have accident and attackers who compromised smart sensors. The Internet of everything is developing a haystack which contains lots of valuable forensics artefacts while identification, collection, preservation and reporting evidences would be a challenge in this environment. This talk would discuss about tools, methods, and techniques to identify, collect and preserve IoT evidences and then elaborates on different challenges that forensics examiners would face in investigating of IoT environments.

CV Relatore: - - -

Torna su ▲

Become Your Own Malware Fortune Teller

Dove / Data e Ora: Meeting Room VII Floor, il giorno 30/05/2016 alle ore 17:00

Relatore: Dr. Ali Dehghantanha (University of Salford UK)

Abstract: Since 1949 that John von Neumann developed the theory of self-reproducing automatons and development of first computer viruses in 1982; malicious programs have always served as a tool for facilitating cyber criminals! A wider range of malwares have been developed since then and we are always hit by new and innovative malwares every day. Being prepared for fighting against future malware and defending the organization network has always been an issue for cyber security teams and forensics investigators. This presentation is elaborating evolvement of different types of malwares namely mobile malwares, macro malwares, and ransomware to reveal the patterns that are keep accruing in development and wide adoption of these programs. Those patterns are then mapped to predict future trends in malicious programs. The talk would elaborate on expected trends in Internet of Things (IoT) malwares, autonomous vehicles malwares and self-learning malicious programs. Finally, the best responses to these malware trends from different stakeholders’ perspective such as law enforcement agents, malware analyst, and incident handlers are discussed and a model for predicting future trends of any malicious program is explained.

CV Relatore: - - -

Torna su ▲

Structure metric learning in prototype-based models and its application for intelligent tutoring

Dove / Data e Ora: aula 2AB45 Torre Archimede, il giorno 21/01/2016 alle ore 10:00

Relatore: Prof. Dr. Barbara Hammer (CITEC centre of excellence Bielefeld University)

Abstract: Prototype-based learning techniques enjoy a wide popularity due to their intuitive training and model interpretability. Applications include biomedical data analysis, image classification, or fault detection in technical systems. Recently, first promising attempts incorporate such models into the domain of intelligent tutoring systems (ITS): in a nutshell, ITSs provide automated, personalised feedback to learners when performing some learning task such as learning how to program. Here a challenge is to avoid time-consuming expert generation of how to provide such feedback; machine learning technology offers promising ways to automate this process, specifically, prototype-based methods enable an automatic feedback generation by highlighting prototype solutions given a learner solution. This strategy relies on the core property of such models that they represent data in terms of typical representatives. Within the talk, we will mainly focus on modern variants of so-called learning vector quantization (LVQ) due to their strong learning theoretical background and exact mathematical derivative from explicit cost functions. The use of LVQ in ITSs faces two challenges: 1) Data are typically non-vectorial, e.g. structured data such as sequences are present; since classical LVQ models have been designed for euclidean vectors only, the question is how to extend LVQ technology towards non-vectorial data. We will present relational extensions of LVQ technology which enable its use for proximity data as provided by structure metrics such as alignment in a very generic way. 2) Structure metrics crucially depend on model parameters such as the scoring function, and their optimum choice is not clear. Still, the accuracy of such models crucially depends on a correct choice of these metric parameters. We will present recent results which allow to adjust structure metric parameters autonomously based on the given data and learning task only.

CV Relatore: http://www.techfak.uni-bielefeld.de/~bhammer/

Torna su ▲

Using Mobile Sensors for Estimating Citywide Pollution Levels

Dove / Data e Ora: aula 1BC50 Torre Archimede, il giorno 20/01/2016 alle ore 10:30

Relatore: Prof. Carlos T. Calafate (Technical University of Valencia (UPV), Spain)

Abstract: Mobile sensing is becoming the best option to monitor our environment due to its ease of use, high flexibility, and low price. In this talk we show the evolution of a project starting from an initial validation procedure using an analytical/simulation-based approach, and then moving towards a real mobile sensing architecture (EcoSensor) able to monitor different air pollutants using low-end sensors. Our architecture is composed of three different modules: a mobile sensor for monitoring environment pollutants, an Android-based device for transferring the gathered data to a central server, and a central processing server for analyzing the collected data through spatial interpolation techniques and generate pollution distribution maps. Besides presenting the architecture itself, we analyze different issues related to the monitoring process: (i) Filtering captured data to reduce the variability of consecutive measurements; (ii) Converting the sensor output to actual pollution levels; (iii) Reducing the temporal variations produced by mobile sensing process; and (iv) Applying interpolation techniques for creating detailed pollution maps. In addition, we study the best strategy to use mobile sensors by first determining the influence of sensor orientation on the captured values, and then analyzing the influence of time and space sampling in the interpolation process. Finally, we detail how the project is evolving towards UAV-based solutions for automated monitoring in rural & adverse ground mobility environments.

CV Relatore: Carlos T. Calafate is an associate professor in the Department of Computer Engineering at the Technical University of Valencia (UPV) in Spain. He graduated with honours in Electrical and Computer Engineering at the University of Oporto (Portugal) in 2001. He received his Ph.D. degree in Informatics from the Technical University of Valencia in 2006, where he has worked since 2002. His research interests include ad-hoc and vehicular networks, mobile applications, QoS, network protocols, video streaming, and network security. To date he has published more than 270 articles, several of them in top conferences and journals. Currently he is leading the Smart@CarPhone, a 3-year project financed by the Spanish government.

Torna su ▲

ShareLaTeX e CfL

Dove / Data e Ora: aula 2BC30 Torre Archimede, il giorno 30/11/2015 alle ore 12:00

Relatore: Luca Tronchin

Abstract: ShareLatex e' uno strumento software per la scrittura collaborativa di testi latex, accessibile da un comune browser internet, che sta diffondendosi nel mondo accademico. Ne verra' fatta una presentazione generale lato utente e poi una piu' tecnica, relativa all'attivita' di inserimento di CfL come servizio aggiuntivo disponibile dall'interfaccia utente di ShareLatex. CfL e' un progetto di ''Computing from Latex'' ovvero di calcolo a partire da un problema matematico scritto in un testo latex: il problema viene riconosciuto dal parser di CfL ed un solutore numerico viene generato in uno script Python, a sua volta eseguito; i risultati vengono restituiti all'interno del testo latex che viene infine compilato in PDF. Verranno presentati alcuni esempi relativi allo stato dell'arte di CfL ed alcune applicazioni in ambito ingegneristico ed educational. Il progetto e' aperto ad ulteriori sviluppi.

CV Relatore: - - -

Torna su ▲

Embedded security in the wild: SCA on small devices and IoT technologies

Dove / Data e Ora: 1AD100, il giorno 21/10/2015 alle ore 10:00

Relatore: Lejla Batina (Radboud University, NL)

Abstract: In this talk we first give an overview of side-channel attacks on embedded devices and we discuss some recent developments in this area. In the second part of the talk we survey some prominent solutions for privacy-friendly RFID identification protocols and discuss their properties and hardware requirements.

CV Relatore: - - -

Torna su ▲

Boten ELISA: A Novel Approach for Botnet C&C in Online Social Networks

Dove / Data e Ora: sala riunioni IV piano, il giorno 24/09/2015 alle ore 13:00

Relatore: Daniele Lain (studente magistrale Universita' di Padova)

Abstract: The Command and Control (C&C) channel of modern botnets is migrating from traditional centralized solutions (such as the ones based on Internet Relay Chat and Hyper Text Transfer Protocol), towards new decentralized approaches. As an example, in order to conceal their traffic and avoid blacklisting mechanisms, recent C&C channels use peer-to-peer networks or abuse popular Online Social Networks (OSNs). A key reason for this paradigm shift is that current detection systems become quite effective in detecting centralized C&C. In this talk, we introduce the evolution of C&C channels and modern detection systems. We then present ELISA (Elusive Social Army), a novel type of botnet that conceals C&C information using OSNs accounts of unaware users. In particular, ELISA exploits in a opportunistic way the messages that users exchange through the OSN. We show that several popular social networks can be maliciously exploited to run this type of botnet, and we discuss why current traffic analysis systems cannot detect ELISA. This work will be also presented at the 2015 IEEE Conference on Communications and Network Security (CNS 2015).

CV Relatore: - - -

Torna su ▲

GNSS Spoofing Attacks and Countermeasures

Dove / Data e Ora: aula 2BC30 Torre Archimede, il giorno 17/09/2015 alle ore 10:00

Relatore: Christina Popper (Ruhr-University Bochum (RUB), Germany)

Abstract: In this talk we will review techniques that allow to identify spoofing attacks on Global Navigation Satellite Systems (GNSS), such as GPS. The specific setup of these systems - in particular the reliance on one-way communication - ease attacks, but make them also particularly interesting for security research. I will present different attack models and scenarios as well as detection solutions that take fundamental principles into account.

CV Relatore: Christina Pöpper is Assistant Professor and head of the Information Security Group at Ruhr-University Bochum (RUB), Germany. She is also a member of HGI, the Horst-Görtz-Institute for IT-Security at RUB. Before joining RUB, she worked as a postdoctoral researcher at the Institute of Information Security at ETH Zurich, where she obtained a Ph.D. in Computer Science in 2011. Prior to that, she worked at the European Space Agency (ESA) and received her Dipl.-Ing. Degree in Computer Science from ETH Zurich in 2005. Her research interests cover various IT-security topics, with a focus on communication, wireless, and protocol-layer security.

Torna su ▲

Data Deduplication and its Security Risks

Dove / Data e Ora: aula 2BC30, il giorno 10/09/2015 alle ore 13:30

Relatore: Chia-Mu Yu (Yuan Ze University, Taiwan)

Abstract: Cloud storage such as Dropbox and Bitcasa is one of the most popular cloud services. Currently, with the prevalence of mobile cloud computing, users can even collaboratively edit the newest version of documents and synchronize the newest files on their smart mobile devices. A remarkable feature of the current cloud storages is their virtually infinite storage. To support the unlimited storage, the cloud storage provider uses data deduplication technique to reduce the data to be stored and therefore reduce the storage expense. Moreover, the use of data deduplication also helps significantly reduce the need of bandwidth and therefore improve the user experience. Nevertheless, in spite of the above benefits, the data deduplication has its inherent security weaknesses. For example, the adversary may have an unauthorized file downloading via the file hash only. In this talk, we will introduce the inherent security risks of data deduplication, review previous solutions, identify their performance weaknesses, and raise potential countermeasures.

CV Relatore: Chia-Mu Yu received his Ph.D. degree from National Taiwan University in 2012. He was a research assistant in the Institute of Information Science, Academia Sinica, Taipei, Taiwan from 2005 to 2010. He was a visiting scholar at Harvard University (Sep 2010 - Sep 2011), a visiting scholar at Imperial College London (Jan 2012 - Sep 2012), a postdoc researcher at IBM Thomas J. Watson Research Center (Oct 2012 - Jul 2013), and a visiting professor at Waseda University (Feb 2015 - Mar 2015). He is currently an Associate Editor of IEEE Access, Associate Editor of Security and Communication Networks, and assistant professor at Department of Computer Science and Engineering, Yuan Ze University, Taiwan. He received Excellent Junior Research Investigator Grant from Ministry of Science and Technology, Taiwan, in 2015. His research interests include cloud security, privacy preservation techniques, botnet/APT detection, and cryptography.

Torna su ▲

Formal Security Analysis of the MaCAN Protocol

Dove / Data e Ora: aula 1BC50, il giorno 11/09/2015 alle ore 13:30

Relatore: Alessandro Bruni

Abstract: Embedded real-time network protocols such as the CAN bus cannot rely on off-the-shelf schemes for authentication, because of the bandwidth limitations imposed by the network. As a result, both academia and industry have proposed custom protocols that meet such constraints, with solutions that may be deemed insecure if considered out of context. MaCAN is one such compatible authentication protocol, proposed by Volkswagen Research and a strong candidate for being adopted by the automotive industry. In this work we formally analyse MaCAN with ProVerif, an automated protocol verifier. Our formal analysis identifies two flaws in the original protocol: one creates unavailability concerns during key establishment, and the other allows re-using authenticated signals for different purposes. We propose and analyse a modification that improves its behaviour while fitting the constraints of CAN bus. Although the revised scheme improves the situation, it is still not completely secure. We argue that the modified protocol makes a good compromise between the desire to secure automotive systems and the limitations of CAN networks, and we discuss the limitations of the analysis tool in analysing this case study, showing an extension of the language that overcomes them.

CV Relatore: Alessandro Bruni is a PhD student at the Technical University of Denmark (DTU), and received his MS and BS from Università degli Studi di Padova. His research interests span across formal verification, model checking, security protocols and machine learning.

Torna su ▲

How to Generate Security Cameras: Towards Defence Generation for Socio-Technical Systems

Dove / Data e Ora: aula 2AB45, il giorno 14/07/2015 alle ore 14:00

Relatore: Olga Gadyatskaya (University of Luxembourg)

Abstract: Recently security researchers have started to look into automated generation of attack trees from socio-technical system models. The obvious next step in this trend of automated risk analysis is automating the selection of security controls to treat the detected threats. However, the existing socio-technical models are too abstract to represent all security controls recommended by practitioners and standards. We propose an attack-defence model, consisting of a set of attack-defence bundles, to be generated and maintained with the socio-technical model. The attack-defence bundles can be used to synthesise attack-defence trees directly from the model to offer basic attack-defence analysis, but they can also be used to select and maintain the security controls that cannot be handled by the model itself. In the talk we will review the concepts of socio-technical models and automated generation of attacks, present the attack-defence model, and discuss the current challenges in the automated risk analysis.

CV Relatore: Olga Gadyatskaya is a Research Associate in the Interdisciplinary Centre for Security, Reliability and Trust (SnT) at the University of Luxembourg. Prior to joining SnT in 2014, she was a post-doctoral researcher at the University of Trento, Italy. She received her PhD in Mathematics at Novosibirsk State University, Russia, in 2008. Her current research spans from mobile systems security to security and risk evaluation for socio-technical systems.

Torna su ▲

StaDynA: Addressing the Problem of Dynamic Code Updates in the Security Analysis of Android Applications

Dove / Data e Ora: aula 2AB45, il giorno 14/07/2015 alle ore 10:30

Relatore: Yury Zhauniarovich (University of Trento)

Abstract: Static analysis of Android applications can be hindered by the presence of the popular dynamic code update techniques: dynamic class loading and reflection. Recent Android malware samples do actually use these mechanisms to conceal their malicious behavior from static analyzers. These techniques defuse even the most recent static analyzers that usually operate under the ``closed world'' assumption (the targets of reflective calls can be resolved at analysis time; only classes reachable from the class path at analysis time are used at runtime). In this work we proposed the solution that allows existing static analyzers to remove this assumption. This is achieved by combining static and dynamic analysis of applications in order to reveal the hidden/updated behavior and extend static analysis results with this information. In this presentation we will describe design, implementation and preliminary evaluation results of our solution called StaDynA.

CV Relatore: Yury Zhauniarovich is a postdoctoral researcher at the University of Trento (Italy) in Security Research Group. He earned his M.Sc. degree in Computer Science from the Belarusian State University in 2007. From 2007 till 2009, he worked as a SAP Consultant at Itransition. In April 2014, he received his Ph.D. degree in Information and Communication Technology from the University of Trento. His research interests include design, implementation and evaluation of security enhancements of mobile operating systems, runtime security, smartphone applications security and mobile malware.

Torna su ▲

Recent Advances in Simulation Technology and Serious Games – Applications from METU Game Technologies Master’s Program

Dove / Data e Ora: 1BC45, il giorno 25/06/2015 alle ore 14:00

Relatore: Dr. Elif Surer (Informatics Institute, METU (Ankara, Turkey))

Abstract: In this talk, recent tools and approaches that are used in the domains of simulation technology and serious games will be introduced. Applications and theses from Middle East Technical University Game Technologies Program regarding flight, cloth simulations, military applications, motion capture and education will also be presented in detail.

CV Relatore: - - -

Torna su ▲

Model Selection and Error Estimation in Learning from Empirical Data

Dove / Data e Ora: Aula 1BC50 Torre Archimede, il giorno 26/06/2015 alle ore 14:30

Relatore: Luca Oneto (SmartLab, DITEN, University of Genoa)

Abstract: In the Supervised Learning framework, a model is built by exploiting the available observations through a Learning Algorithm that is able to capture the information hidden in the data. Model Selection addresses the problem of tuning a Learning Algorithm to the available data in order to reduce the Generalization (True) Error of the final model. This problem affects most of the algorithms because, in general, their effectiveness is controlled by one or more hyperparameters which must be tuned during the learning process for achieving optimal performances. Associated to the issues of Model Selection we find the problem of estimating the True Error of a classifier: in fact, the main objective of building an optimal classifier is to choose the parameters and hyperparameters that minimize its True Error and compute an estimate of this value for predicting the classification performance on future data. Unfortunately, despite the large amount of work done on this important topic, the problem of Model Selection and Error Estimation for a Learning Algorithm is still open and the focus of extensive research. The purpose of this seminar is to give an overview of the problem of Model Selection and Error Estimation. We will start from the seminal works of the 80s until the most recent results on this topic. Finally we will discuss future directions of this multidisciplinary field of research

CV Relatore: Luca Oneto was born in Rapallo, Italy in 1986. He is currently a Researcher at University of Genoa with particular interests in Machine Learning, Statistical Learning Theory and Data Mining. He re- ceived his Bachelor Degree in Electronic Engineering at the University of Genoa, Italy in 2008. He subsequently started his master studies in Electronic Engineering in the same university with focus in Intelligent Systems and Statistics. After receiving his MSc Degree in 2010, he started to work as a consultant for the DITEN and DIBE Departments at University of Genoa, together with other con- sultant activities for Mac96 and Ansaldo STS in the context of many European Projects. In 2014 he received his PhD in School of Sciences and Technologies for Knowledge and Information Retrieval (University of Genoa) with the thesis ”Learning Based On Empirical Data”. Today he works as a consultant and teaches in many BSc and MSc courses at University of Genoa as a Researcher.

Torna su ▲

Impact of Country-scale Internet Disconnection on Structured and Social P2P Overlays

Dove / Data e Ora: 1BC45, il giorno 10/06/2015 alle ore 14:00

Relatore: Ding Ding (University of Padua)

Abstract: Peer-to-peer systems are resilient in the presence of churn and uncorrelated failures. However, their behavior in extreme scenarios where massive correlated failures occur is not well-studied. Yet, there have been examples of situations where a country-scale fraction of Internet users have been disconnected from the rest of the network—for instance, when a government cuts connectivity to the outside world as a mechanism for suppression of uprisings. In this paper, we consider the effect of such partitions on topology and routing of structured and social-based unstructured P2P overlays, including a novel social-aware overlay. In particular, we consider nodes within a relatively small fraction of the network (2.5% or fewer Internet users), and study whether users can communicate with their (n-hop away) social neighbors in a peer-to-peer fashion after the partition. We perform an extensive simulation-based analysis to assess the probability for these communications to be possible. In our analysis, we consider both real and synthetic datasets of online social networks. Our results show that structured P2P overlay routability is severely hampered by country-scale partition events. In addition, the proposed social-based unstructured overlay network provides improved routability while maintaining a smaller number of links.

CV Relatore: - - -

Torna su ▲

Fighting Authorship Linkability with Crowdsourcing

Dove / Data e Ora: 1BC 50 , il giorno 09/06/2015 alle ore 11:00

Relatore: Prof. Gene Tsudik (University of California, Irvine)

Abstract: Massive amounts of contributed content -- including traditional literature, blogs, music, videos, reviews and tweets -- are available on the Internet today, with authors numbering in many millions. Textual information, such as product or service reviews, is an important and increasingly popular type of content that is being used as a foundation of many trendy community-based reviewing sites, such as TripAdvisor and Yelp. Some recent results have shown that, due partly to their specialized/topical nature, sets of reviews authored by the same person are readily linkable based on simple stylometric features. In practice, this means that individuals who author more than a few reviews under different accounts (whether within one site or across multiple sites) can be linked, which represents a significant loss of privacy. In this work, we start by showing that the problem is actually worse than previously believed. We then explore ways to mitigate authorship linkability in community-based reviewing. We first attempt to harness the global power of crowdsourcing by engaging random strangers into the process of re-writing reviews. As our empirical results (obtained from Amazon Mechanical Turk) clearly demonstrate, crowdsourcing yields impressively sensible reviews that reflect sufficiently different stylometric characteristics such that prior stylometric linkability techniques become largely ineffective. We also consider using machine translation to automatically re-write reviews. Contrary to what was previously believed, our results show that translation decreases authorship linkability as the number of intermediate languages grows. Finally, we explore the combination of crowdsourcing and machine translation and report on results.

CV Relatore: - - -

Torna su ▲

Violating Consumer Anonymity: Geo-locating Nodes in Named Data Networking

Dove / Data e Ora: 1BC45, il giorno 28/05/2015 alle ore 13:00

Relatore: Alberto Compagno (Sapienza University of Rome)

Abstract: Named Data Networking (NDN) is an instance of information-centric network architecture designed as a candidate replacement for the current IP-based Internet. It emphasizes efficient content distribution, achieved via in-network caching and collapsing of closely-spaced content requests. NDN also offers strong security and explicitly decouples content from entities that distribute it. NDN is widely assumed to provide better privacy than IP, mainly because NDN packets lack source and destination addresses. In this paper, we show that this assumption does not hold in practice. In particular, we present several algorithms that help locate consumers by taking advantage of NDN router-side content caching. We use simulations to evaluate these algorithms on a large and realistic topology, and validate the results on the official NDN testbed. Beyond locating consumers, proposed techniques can also be used to detect eavesdroppers.

CV Relatore: - - -

Torna su ▲

Bitcoin Forensics - Where are my coins?

Dove / Data e Ora: aula 2BC30, il giorno 26/05/2015 alle ore 17:30

Relatore: Matteo Brunati (CyBrain srl)

Abstract: Bitcoin is a technology and a social phenomenon which raised lots of public interest after the seizure of Silkroad website since winter 2013/2014. Even though the public interest seems diminished during 2015, analyzing the Bitcoin network we see that transactions numbers and volumes actually increased since last year, thus confirming the thesis of many Police Forces worldwide which see a usage increase of this technology between criminals. During this seminar we will analyze the Bitcoin technology from a Digital Forensics point of view, looking at Bitcoin wallets, transactions, software and network in order to understand and investigate its use. We will cover also an introduction to Bitcoin antiforensics, pinpointing some techniques to improve Bitcoin users privacy.

CV Relatore: - - -

Torna su ▲

Proof, Secrets, and Computation

Dove / Data e Ora: Palazzo del Bo - Aula Magna Galileo Galilei, il giorno 25/05/2015 alle ore 09:30

Relatore: Silvio Micali (MIT Boston)

Abstract: We show how Theory of Computation has revolutionized our millenary notion of a proof, revealing its unexpected applications to our new digital world. In particular, we shall demonstrate how interaction can make proofs much easier to verify, dramatically limit the amount of knowledge released, and yield the most secure identification schemes to date.

CV Relatore: Silvio Micali has received his Laurea in Mathematics from the University of Rome, and his PhD in Computer Science from the University of California at Berkeley. Since 1983 he has been on the MIT faculty, in Electrical Engineering and Computer Science Department, where he is Ford Professor of Engineering and Associate Department Head. Silvio's research interests are cryptography, zero knowledge, pseudo-random generation, secure protocols, and mechanism design. Silvio has received the Turing Award (in computer science), the Gödel Prize (in theoretical computer science), and the RSA prize (in cryptography). He is a member of the National Academy of Sciences, the National Academy of Engineering, and the American Academy of Arts and Sciences.

Torna su ▲

Neurodynamic Optimization Approaches to Parallel Data Selection in the Era of Big Data

Dove / Data e Ora: aula 2AB40 Torre Archimede, il giorno 29/05/2015 alle ore 14:30

Relatore: Jun Wang (The Chinese University of Hong Kong)

Abstract: In the present information era, huge amount of data to be processed daily. In contrast of conventional sequential data processing techniques, parallel data processing approaches can expedite the processes and more efficiently deal with big data. In the last few decades, neural computation emerged as a popular area for parallel and distributed data processing. The data processing applications of neural computation included, but not limited to, data sorting, data selection, data mining, data fusion, and data reconciliation. In this talk, neurodynamic approaches to parallel data processing will be introduced, reviewed, and compared. In particular, my talk will compare several mathematical problem formulations of well-known multiple winners-take-all problem and present several recurrent neural networks with reducing model complexity. Finally, the best one with the simplest model complexity and maximum computational efficiency will be highlighted.  Analytical and Monte Carlo simulation results will be shown to demonstrate the computing characteristics and performance of the continuous-time and discrete-time models. The applications to parallel sorting, rank-order filtering, and data retrieval will be also discussed.

CV Relatore: Jun Wang is a Professor and the Director of the Computational Intelligence Laboratory in the Department of Mechanical and Automation Engineering at the Chinese University of Hong Kong. Prior to this position, he held various academic positions at Dalian University of Technology, Case Western Reserve University, and University of North Dakota. He also held various short-term visiting positions at USAF Armstrong Laboratory (1995), RIKEN Brain Science Institute (2001), Universite Catholique de Louvain (2001), Chinese Academy of Sciences (2002), Huazhong University of Science and Technology (2006–2007), and Shanghai Jiao Tong University (2008-2011) as a Changjiang Chair Professor. Since 2011, he is a National Thousand-Talent Chair Professor at Dalian University of Technology on a part-time basis. He received a B.S. degree in electrical engineering and an M.S. degree in systems engineering from Dalian University of Technology, Dalian, China. He received his Ph.D. degree in systems engineering from Case Western Reserve University, Cleveland, Ohio, USA. His current research interests include neural networks and their applications. He published over 170 journal papers, 15 book chapters, 11 edited books, and numerous conference papers in these areas. He is the Editor-in-Chief of the IEEE Transactions on Cybernetics since 2014 and a member of the editorial board of Neural Networks since 2012. He also served as an Associate Editor of the IEEE Transactions on Neural Networks (1999-2009), IEEE Transactions on Cybernetics and its predecessor (2003-2013), and IEEE Transactions on Systems, Man, and Cybernetics – Part C (2002–2005), as a member of the editorial advisory board of International Journal of Neural Systems (2006-2013), as a guest editor of special issues of European Journal of Operational Research (1996), International Journal of Neural Systems (2007), Neurocomputing (2008, 2014), and International Journal of Fuzzy Systems (2010, 2011). He was an organizer of several international conferences such as the General Chair of the 13th International Conference on Neural Information Processing (2006) and the 2008 IEEE World Congress on Computational Intelligence, and a Program Chair of the IEEE International Conference on Systems, Man, and Cybernetics (2012). He has been an IEEE Computational Intelligence Society Distinguished Lecturer (2010-2012, 2014-2016). In addition, he served as President of Asia Pacific Neural Network Assembly (APNNA) in 2006 and many organizations such as IEEE Fellow Committee (2011-2012); IEEE Computational Intelligence Society Awards Committee (2008, 2012, 2014), IEEE Systems, Man, and Cybernetics Society Board of Directors (2013-2015), He is an IEEE Fellow, IAPR Fellow, and a recipient of an IEEE Transactions on Neural Networks Outstanding Paper Award and APNNA Outstanding Achievement Award in 2011, Natural Science Awards from Shanghai Municipal Government (2009) and Ministry of Education of China (2011), and Neural Networks Pioneer Award from IEEE Computational Intelligence Society (2014), among others.

Torna su ▲

Intelligent Transportation Systems and Smart Cities: Overview of the Research activities of the Grupo de Redes de Computadores (GRC), Universitat Politecnica de Valencia

Dove / Data e Ora: Room 1BC/45 Torre Archimede, il giorno 21/01/2015 alle ore 11:30

Relatore: Prof. Pietro Manzoni (Universitat Politecnica de Valencia)

Abstract: Wireless communication for intelligent transportation systems (ITSs) and Smart Cities is a promising technology to improve driving safety, reduce traffic congestion and support information services in vehicles. During recent ITS development, transportation telematics techniques have exhibited much progress, e.g., interaction between automobiles and the infrastructure for delivering services such as road-side assistance, automatic crash notification, concierge assistance and vehicle condition reports. This presentation provides an overview of the research activities that are being carried out in the Networking Group of the Universitat Politecnica de Valencia on this topic. The results and the future work from the cooperation with an industrial partner will also be described. This seminar is organized by Prof. Claudio Palazzi

CV Relatore: Pietro Manzoni received the MS degree in computer science from the Universita' degli Studi of Milan, Italy, in 1989, and the PhD degree in computer science from the Politecnico di Milano, Italy, in 1995. He is currently a full professor of computer science at the Universitat Politecnica de Valencia, Spain in the Department of Computer Engineering. His research activity is related to mobile wireless data systems design, modelling, and implementation, particularly oriented to Intelligent Transport Systems and Smart Cities. He published more than 200 scientific papers, 49 of them in international journals with impact; his H-index is 25 according to Google Scholar. He has been actively involved in the organization or in the technical committee of various scientific conference and journals. He collaborates with various international academic and industrial research centres. He is a member of the IEEE.

Torna su ▲

Cyber-Physical Security in Future Cities

Dove / Data e Ora: Aula 1BC45, il giorno 03/11/2014 alle ore 14:30

Relatore: David Yau (Singapore University of Technology and Design)

Abstract: Digital control and communication are being used to modernize urban infrastructures, such as electrical grids and water systems, to meet the increasing demands of future cities for size, performance, and sustainability. While the added intelligence provides desirable performance features, it also adds to the system's complexity, which raises equally critical concerns for safety and security. In this talk, I will overview research that analyzes cyber-physical system (CPS) vulnerabilities such as real-time consumer pricing in emerging smart-grid demand response. I will also propose a design methodology to monitor the (not fully trustworthy) high performance operation of a smart grid, but assure its fallback to a simple and safe operation mode when the system drifts too close to unsafety.

CV Relatore: David Yau received the B.Sc. (first class honors) from the Chinese University of Hong Kong, and M.S. and Ph.D. from the University of Texas at Austin, all in computer science. He has been Professor of Informations Systems Technology and Design at SUTD since 2013. Since 2010, he has been Distinguished Scientist at the Advanced Digital Sciences Center, Singapore, where he leads the Smart Grid IT research program. Prior to Singapore, he was Associate Professor of Computer Science at Purdue University (West Lafayette), USA. David’s research interests are in network protocol design and implementation, CPS security and privacy, quality of service, network incentives, and wireless and sensor networks. He received a CAREER award from the U.S. National Science Foundation. He was also the recipient of an IBM PhD Fellowship. He won Best Paper award from the 2010 IEEE International Conference on Multi-sensor Fusion and Integration (MFI). His papers in 2008 IEEE Int'l Conf. Mobile Ad-hoc and Sensor Systems (MASS), 2013 IEEE Int'l Conf. Pervasive Computing and Communications (PerCom), 2013 IEEE Int'l Conf. Cyber-Physical Systems, Networks, and Applications (CPSNA), and 2013 ACM BuildSys were Best Paper finalists.

Torna su ▲

Staying Alive: System Design for Self-Sufficient Sensor Networks

Dove / Data e Ora: Aula 1BC/50 - Dipartimento di Matematica - Torre di Archimede, il giorno 04/11/2014 alle ore 11:30

Relatore: Michele Rossi (DEI - Universita' degli Studi di Padova)

Abstract: Self-sustainability is a crucial step for modern wireless sensor networks (WSN). In this talk, I will offer an original mathematical framework for autonomous WSN powered by renewable energy sources. At first, the problem at stake will be decomposed into two nested optimization steps: the inner step characterizes the optimal network operating point subject to an average energy consumption constraint, while the outer step provides online energy management policies that make the system energetically self-sufficient in the presence of intermittent (Markov-modulated) energy sources. This framework sheds new light into the design of pragmatic schemes for the control of energy harvesting sensor networks and permits to gauge the impact of key sensor network parameters, such as the battery capacity, the harvester size, the information transmission rate and the radio duty cycle. The obtained (online) energy management policies are finally evaluated considering real solar radiation traces, validating them against state of the art solutions and describing the impact of relevant design choices in terms of achievable network throughput and battery level dynamics.

CV Relatore: Michele Rossi is Assistant Professor of wireless networking at the Department of Engineering (DEI), University of Padova. He received the MS degree in Electrical Engineering and the Ph.D. in Telecommunications Engineering from the University of Ferrara (Italy) in 2000 and 2004, respectively. In 2003 he has been with the Center for Wireless Communications (CWC) at the University of California, San Diego, where he performed research on Wireless Sensor Networks (WSNs) under the supervision of Prof. Ramesh R. Rao. Since november 2005 he has been with the department of Information Engineering @ the University of Padova. Dr. Rossi is actively involved in local as well as EU funded projects, is author of about 100 papers published in peer reviewed International journals, book chapters and conferences and is the recipient of four IEEE Best Paper Awards. In 2010-2014 he has been a Marie Curie fellow within the FP7 SWAP project (on energy harvesting sensor networks). Dr. Rossi is a Senior Member of the IEEE and serves as Associate Editor for the IEEE Transactions on Wireless Communications. His research interests include: the design, the stochastic modeling and the optimization of wireless systems, the use of application layer coding, spatio-temporal compression for WSNs, protocol design for energy harvesting WSNs and Internet of Things, telecommunication technology for smart energy grids.

Torna su ▲

Security and Privacy in Molecular Communications and Nanonetworks

Dove / Data e Ora: aula 2BC30 Torre Archimede, il giorno 31/10/2014 alle ore 09:30

Relatore: Alberto Giaretta (MSc student in Computer Science - Universita' di Padova)

Abstract: Nanotechnology might be the silver bullet in the future of several sectors such as the biomedical and military ones. Molecular Communication paradigm has some distinct characteristics from traditional communication paradigm and these differences are of primary importance in order to achieve reliable and efficient nanonetworks. Being a different communication paradigm, Molecular Communication raises open issues about security and data privacy that are not easily solvable with standard approaches. In this context, this presentation will give an overview about: Nanonetworks, Molecular Communication paradigm, and their security issues. The content of this presentation is based on the following papers: 1. Valeria Loscri, Cesar Marchal, Nathalie Mitton, Giancarlo Fortino, Athanasios V. Vasilakos, ''Security and Privacy in Molecular Communication and Networking: Opportunities and Challenges'', IEEE Transactions on Nanobioscience, accepted, 2014 2. Ian F. Akyildiz, Fernando Brunetti, Cristina Blázquez, ''Nanonetworks: A new communication paradigm'', Computer Networks, 52, 2260–2279, 2008. 3. Sasitharan Balasubramaniam, Pietro Lio’, ''Multi-hop Conjugation based Bacteria Nanonetworks'', IEEE Transactions on NanoBioscience, vol. 12, no. 1, pp.47-59, March 2013. 4. Tadashi Nakano and Athanasios V. Vasilakos, Guest Editorial Special Section on Molecular Communication. IEEE Transactions on NanoBioscience, Vol. 13, no. 3, 2014.

CV Relatore: - - -

Torna su ▲

3D Shape matching by bag-of-feature descriptors

Dove / Data e Ora: aula 2AB40 Torre Archimede, il giorno 30/10/2014 alle ore 11:30

Relatore: Umberto Castellani (Universita di Verona)

Abstract: 3D shape matching is very important in a wide variety of fields such as computer graphics, computer vision, and medical image analysis with applications like object recognition, automatic medical diagnosis, and content-based shape retrieval. Recent methods are based on the so called Bag-of-Features (BoF) paradigm commonly used in text and image retrieval by first computing local shape descriptors, and then representing each shape in a ‘geometric dictionary’ using vector quantization. In this talk the matching approach based on the BoF framework will be introduced by proposing several variants and advanced aspects of the involved pipeline (i.e., supervised and non-supervised). Finally, some case studies will be reported for 3D shape retrieval and shape-based medical image classification.

CV Relatore: Umberto Castellani is Ricercatore (i.e., Assistant Professor) of Department of Computer Science at University of Verona. He received his Dottorato di Ricerca (PhD) in Computer Science from the University of Verona in 2003 working on 3D data modelling and reconstruction. He held visiting research positions at Edinburgh University (UK), Universite' Blaise Pascal (France), Michigan State University (USA), Universite' D'Auvergne (France), Italian Institute of Technology (IIT), and University College London (UK). His research is focused on 3D data processing, statistical learning and medical image analysis. He has co-authored several papers published in leading conference proceedings and journals. He is teaching Computer Vision at the Computer Science Department, and Multimedia at the Department of Filologia, Letteratura e Linguistica at the University of Verona. He is member of Eurographics, IAPR, MICCAI and IEEE.

Torna su ▲