 |
Memorization in Machine Learning
Abstract:
In this talk, I will present leave-one-out distinguishability as a novel approach to measuring data memorization in machine learning models. This technique quantifies how much a model's predictions depend on the inclusion or exclusion of specific training data points, offering a clear indicator of memorization. I will also discuss state-of-the-art membership inference attacks, which are statistical tests designed to evaluate the degree to which a model may leak information about its memorized training data. These attacks help expose potential privacy risks, demonstrating how adversaries could infer whether certain data points were part of the model's training set.
Biography:
Reza Shokri is a Dean's Chair Associate Professor of Computer Science at NUS, and an Asian Young Scientist Fellow. His research focuses on data privacy and trustworthy machine learning. He is an active member of the privacy and security research community, and has served on the program committees of IEEE S&P, ACM CCS, USENIX Security, NDSS, and PETS, as well as major award committees on privacy. He is a recipient of the IEEE Security and Privacy (S&P) Test-of-Time Award 2021, and the Caspar Bowden Award for Outstanding Research in Privacy Enhancing Technologies in 2018 for his work on quantifying privacy, the Best Paper Award at ACM Conference on Fairness, Accountability, and Transparency (FAccT) 2023 for his work on analyzing fairness in machine learning. He has also received the Intel's 2023 Outstanding Researcher Award, VMWare Early Career Faculty Award 2021, Meta (Facebook) Faculty Research Award 2021, Google Faculty Award on Data Protection and Privacy 2021, and Intel Faculty Research Award (Private AI Collaborative Research Institute) 2021-2023, and NUS Early Career Research Award 2019. He obtained his PhD from EPFL.
|
 |
Lattice-Based Optimization Techniques and Algorithms and Evaluations
Abstract:
I will be talking about my various contributions to the secure and efficient implementation techniques and optimizations of lattice-based cryptographic primitives and protocols across multiple programming languages, CPUs, GPUs, and constrained devices.
Biography:
Amin Sakzad has got a Ph.D. degree in Applied Mathematics from Amirkabir University of Technology (AUT), Tehran, Iran, 2011. He was a research visitor and a lecturer at Carleton University, Ottawa, Canada, in 2010. He was a research lecturer at AUT in 2011. Starting from Jan. 2012, he was a research fellow at Software Defined Telecommunications (SDT) Laboratory in the Department of Electrical and Computer Systems Engineering at Monash University under supervision of Prof. Emanuele Viterbo. From Feb. 2015 to April 2017, he was a research fellow at Clayton School of Information Technology at Monash University under supervision of Associate Professor Ron Steinfeld. As of May 2017, he was a Lecturer at Faculty of Information Technology at Monash University. As of July 2021, Amin was promoted to a Senior Lecturer (Associate Professor) at FIT. Since July 2024, Amin is an Associate Professor at FIT. Dr. Amin Sakzad is mainly interested in applications of Euclidean lattices in cryptography and wireless communications. This includes applications of Algebraic Number Theory, Diophantine Approximation and Finite Fields in physical layer network coding and security, Multiple-Input Multiple-Output (MIMO) channels, lattice-based cryptography, and searchable encryption.
|
.png) |
Zero Knowledge Proofs in Blockchain
Abstract:
In this talk, we will explore the fundamental concept of Zero Knowledge Proofs (ZKPs), a cryptographic technique that allows one party to prove knowledge of a specific piece of information to another party without revealing the information itself. We will delve into two core computation models used in ZKPs: the circuit computation model and the machine computation model. These models form the basis of how ZKPs are designed and implemented. Furthermore, we will examine various real-world applications of ZKPs, with particular emphasis on their critical role in enhancing privacy and security in blockchain systems. Attendees will gain a foundational understanding of ZKPs and insight into how they are applied to diverse scenarios in modern cryptographic systems.
Biography:
José L. Muñoz-Tapia received the M.S. degree in telecommunications engineering and the Ph.D. degree in security engineering, in 1999 and 2003, respectively. He is currently a Researcher with the Information Security Group (ISG) and an Associate Professor with the Department of Network Engineering, Universitat Politècnica de Catalunya (UPC). He is also the Director of the Master Program in Blockchain Technologies, UPC School. He has worked in applied cryptography, network security, and game theory models applied to networks and simulators. His research interest includes distributed ledgers technologies.
|
 |
How the Iranian 7NDP Satisfies Our CyberSecurity Challenges?
Biography:
Rasool Jalili received his B.S. degree in computer science from Ferdowsi University of Mashhad in 1985, and M.S. degree in computer engineering from Sharif University of Technology in 1989. He received his Ph.D. in computer science from University of Sydney, Australia, in 1995. He then joined the Department of Computer Engineering, Sharif University of Technology in 1995. He has published more than 140 papers in international journals and conference proceedings. He is now an associate professor, doing research in the areas of computer dependability and security, access control, distributed systems, and database systems in his Data and Network Security Laboratory (DNSL).
|
.png) |
Fault Attack on Symmetric Ciphers
Abstract:
In the rapidly changing domain of cybersecurity, the integrity of data encryption ensured by symmetric ciphers serves as a crucial primitive in safeguarding sensitive information from unauthorized access and potential breaches. While the evolution of technology has undoubtedly enhanced the capabilities of security schemes, it has also ushered in a new era of sophisticated attack vectors. This talk will focus on fault attacks that target symmetric ciphers, where even the most resilient systems can be compromised due to the subtle interactions between cryptographic algorithms and hardware weaknesses. The talk delves into fault attacks on symmetric ciphers, ranging from well-known Differential Fault Attacks (DFA) to statistical approaches. For DFA, the talk will explore the main idea behind DFA on a block cipher and address the challenge of applying DFA on a nonce-respecting authenticated encryption system, where the adversary cannot rely on the classical DFA assumption. Subsequently, it will discuss statistical approaches and an extension of the relatively new fault model, the Persistent Fault Analysis (PFA), by examining multiple persistent fault models and statistically ineffective PFAs. In the final segment, various countermeasures will be presented and discussed.
Biography:
Nasour Bagheri received the M.S. and Ph.D. degrees in electrical engineering from the Iran University of Science and Technology (IUST), Tehran, Iran, in 2002 and 2010, respectively, after obtaining a B.Sc. degree in electrical engineering from Mazandaran University in 2000. He is currently a Professor in the Electrical Engineering Department at Shahid Rajaee Teacher Training University, Tehran, and the head of the CPS2 laboratory there. Additionally, he is a part-time researcher at the Institute for Research in Fundamental Sciences. He is the author of more than 100 articles on information security and cryptology. His research interests include cryptology, specifically designing and analyzing symmetric schemes such as lightweight ciphers (e.g., block ciphers), hash functions, and authenticated encryption schemes. He also focuses on hardware security, particularly the security of symmetric schemes against side-channel attacks such as fault injection and power analysis, as well as the security of cryptographic protocols for constrained environments like RFID tags and IoT edge devices.
|
View: 349 Time(s) |
Print: 59 Time(s) |
Email: 0 Time(s) |
0 Comment(s)