Speaker: Berk Gulmezoglu, Ph.D. candidate with the Department of Electrical and Computer Engineering at Worcester Polytechnic Institute
Title: Towards Automated Analysis of Microarchitectural Attacks using Machine Learning
Abstract: Microarchitectural attacks have become a serious threat to the architecture industry (Intel, AMD and so on) in the last decade. While there are many ways to exploit the microarchitectural components, in this talk, we will discuss one of the basic methods called Prime and Probe. To demonstrate the real-world applicability of Prime and Probe attack, I will explain how to collect side-channel data to steal an RSA key of a third-party in a public cloud Amazon EC2. In the second part of the talk, we will discuss how to automate the analysis of the side-channel data by implementing various Machine Learning/Deep Learning techniques. I will show the audience a new target for micro-architectural attacks and how these attacks might violate the privacy of computer users. Finally, I will briefly explain a method called “FortuneTeller” which detects microarchitectural attacks using unsupervised Recurrent Neural Networks in the public servers.
Bio: Berk Gulmezoglu is currently a Ph.D. candidate at the Department of Electrical and Computer Engineering in Worcester Polytechnic Institute. His research focuses include microarchitectural attacks and automating the analysis of side-channel data with Machine Learning methods. He discovered several vulnerabilities in Intel microarchitecture and VMware/Xen hypervisors which received an extensive coverage from the news. Currently, he focuses on exploiting new vulnerabilities in AMD architecture and implementing Natural Language Processing techniques to detect the vulnerable software gadgets.
Seminar Host: Degang Chen