Keynote speakers

Stjepan Picek Assistant professor in the Cybersecurity group at TU Delft, The Netherlands

A Decade of Machine Learning in Profiled Side-channel Analysis

Abstract: The growing markets of embedded computing, and especially Internet-of-Things, require large amounts of confidential data to be processed on electronic devices. Cryptographic algorithms are usually implemented as part of those systems and, if not adequately protected, are vulnerable to side-channel analysis (SCA). In SCA, the attacker exploits weaknesses in the physical implementations of cryptographic algorithms. SCA can be divided into profiled and non-profiled analysis. The profiled side-channel analysis considers a scenario where the adversary has control over a device identical to the target device. The adversary then learns statistics from the device under control and tries to match them on other target devices. In the last decade, profiled side-channel analysis based on machine learning proved to be very successful in breaking cryptographic implementations in various settings. Despite successful attacks, even in the presence of countermeasures, there are many open questions. A large part of the research concentrates on improving the performance of attacks. At the same time, little is done to understand them and, even more importantly, use that knowledge to design more secure implementations. In this talk, we start by discussing success stories on machine learning-based side-channel analysis. First, we discuss how to improve deep learning-based SCA performance by using ensembles of neural networks. Next, we concentrate on the pre-processing phase and explore whether the countermeasures could be considered noise and removed before the attack. More precisely, we investigate the performance of denoising autoencoders to remove several types of hiding countermeasures. Finally, we discuss how to use the results of non-profiled methods as a start for the deep learning-based analysis. To be able to run such an analysis, we discuss the iterative deep learning-based SCA framework for public-key cryptography. In the second part of this talk, we concentrate on critical open questions and research directions that still need to be explored. Here, we will discuss questions like the difference between machine learning and SCA metrics, or portability (the differences between the training and testing devices). Finally, we briefly connect the machine learning progress in profiled SCA with developments in other security domains (hardware Trojans, fault injection attacks).   

Bio: 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. Before the assistant professor position, Stjepan was a postdoctoral researcher at MIT, USA, and at KU Leuven, Belgium. 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. 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 and 2021, program committee member and reviewer for several conferences and journals, and a member of several professional societies.

Chuan Li Professor in the School of Mechanical Engineering at the Dongguan University of Technology, China

Deep learning for machinery fault diagnosis

Abstract: Machinery fault diagnosis aims at diagnosing the causes of the degradation of the industrial machineary and assessing the degradation level, with the objective of increasing the system availability and reducing operation and maintenance costs. In the era of Industry 4.0, the increased availability of information from industrial monitored machinery and the grown ability of treating the acquired information by intelligent algorithms have opened the wide doors for the development of advanced diagnosis methods. However, traditional methods highly rely on the design and selection of handcrafted features, which require the knowledge of experts and the use of computationally intensive trial and error approaches. The large amount of available data poses the problem of extracting and selecting the features relevant for the development of the fault diagnosis methods. Deep learning is a promising tool for dealing with this problem. Deep learning refers to a class of methods that are capable of extracting hierarchical representations from huge volumes of large-dimensional data by using neural networks with multiple layers of non-linear transformations. Therefore, they are expected to directly and automatically provide high-level abstractions of the big data available in Industry 4.0, without requiring human-designed and labor-intensive analyses of the data for the extraction of degradation features. This talk aims at contributing to the above scenario by presenting advanced deep learning methods for modelling complex industrial machinery and treating their data, with the objective of diagnosing their failures.   

Bio: Chuan Li (FIET, SMIEEE) is Professor in the School of Mechanical Engineering at the Dongguan University of Technology, China. In 2007 he obtained PhD degree from Chongqing University, China. He was then a Postdoctoral Fellow with the University of Ottawa, Canada; a Research Professor with the Korea University, South Korea; a Senior Research Associate with the City University of Hong Kong, China; a Prometeo Researcher with the Universidad Politecnica Salesiana, Ecuador; and a Visiting Full Professor with the Universidade do Algarve, Portugal. In 2016 he joined in Dongguan University of Technology where he is currently Founding Director of Guangdong Research Center for Intelligent Manufacturing and Health Maintenance, Leader of Guangdong Key Discipline of Mechanical Engineering, Leader of Guangdong University Innovation Team of Intelligent Manufacturing and PHM, and Dongguan Characteristic Talent. His research interests include machine learning related theories and its practical applications to many interdisciplinary domains but not limited to complex systems, internet of things, and prognosis and health management of mechanical systems. Prof. Li has supervised/ co-supervised over 30 postdoctoral fellows and PhD and Master students. His research has resulted in 100+ journal papers (14 ESI highly cited papers), 50+ conference articles, and over 30 USA, Japanese, and Chinese patents. He has H Index 37, I10 Index 87, and total citations more than 4500 (Google). He has received numerous prestigious awards including National Science and Technology Award of China. As the Principal Investigator, Prof. Li has successfully completed over 30 projects from various funding agencies. He has been serving as Director of PHM Committee of China Technology Market Association, and Chair of IEEE RS Chongqing Chapter.

Julien Mairal Research Scientist at Inria Grenoble, France

Generic acceleration schemes for large-scale optimization in machine learning

Abstract: In this talk, we will present a few optimization principles that have been shown to be useful to address large-scale problems in machine learning. We will focus on recent variants of the stochastic gradient descent method that benefit from several acceleration mechanisms such as variance reduction and Nesterov's extrapolation. We will discuss both theoretical results in terms of complexity analysis, and practical deployment of these approaches, demonstrating that even though Nesterov's acceleration method is almost 40 years old, it is still highly relevant today.   

Bio: Julien Mairal received his PhD in 2010 under the supervision of Francis Bach and Jean Ponce from ENS Cachan. Before that he did his undergraduate studies at Ecole Polytechnique from 2002 to 2005 in Paris, then graduated from Telecom ParisTech and ENS Cachan in 2006 with a master degree. After his PhD, Julien Mairal spent two years as a post-doctoral researcher in the statistics department of UC Berkeley, working with Bin Yu, before joining Inria in 2012. In 2016, Julien Mairal received an ERC starting grant to run the project SOLARIS. His interests include machine learning, optimization, statistical signal and image processing, computer vision, Julien Mairal also has collaborations in bioinformatics.

Efstratios Gavves Assistant Professor at University of Amsterdam, Netherlands

The Machine Learning of Time and Applications

Abstract: Visual artificial intelligence automatically interprets what happens in visual data like videos. Today’s research strives with queries like: “Is this person playing basketball?”; “Find the location of the brain stroke”; or “Track the glacier fractures in satellite footage”. All these queries are about visual observations already taken place. Today’s algorithms focus on explaining past visual observations. Naturally, not all queries are about the past: “Will this person draw something in or out of their pocket?”; “Where will the tumour be in 5 seconds given breathing patterns and moving organs?”; or, “How will the glacier fracture given the current motion and melting patterns?”. For these queries and all others, the next generation of visual algorithms must expect what happens next given past visual observations. Visual artificial intelligence must also be able to prevent before the fact, rather than explain only after it. In this talk, I will present my vision on what these algorithms should look like and how we can obtain them. Furthermore, I will present some recent works and applications in this direction within my lab and spinoff.   

Bio: Dr. Efstratios Gavves is an Associate Professor with the University of Amsterdam in the Netherlands and Scientific Director of the QUVA Deep Vision Lab. He is a recipient of the prestigious ERC Career Starting Grant 2020 to research on the Computational Learning of Temporality for spatiotemporal sequences. Also, he is a co-founder of Ellogon.AI, a University spinoff and in collaboration with the Dutch Cancer Institute (NKI), with the mission of using AI for pathology and genomics. He is currently supervising more than 12 Ph.D. and postdoctoral students in projects with the University of Amsterdam, the Dutch Cancer Institute, Ellogon.AI, and BMW. Efstratios has authored several papers in the top Computer Vision and Machine Learning conferences and journals and he is also the author of several patents. Further, Efstratios teaches Deep Learning in the MSc in Artificial Intelligence at the University of Amsterdam. All material is available on the project website, uvadlc.github.io. His research focus is on Temporal Machine Learning and Dynamics, Efficient Computer Vision, and Machine Learning for Oncology.