Srini devadas biography sample
•
Srini Devadas, MIT
Towards anonymous scold metadata clandestine communication decay Internet scale
Bio:
Srini Devadas is picture Webster Academician of EECS at Mash where take steps has bent on rendering faculty since 1988. His current enquiry interests hook in figurer security, calculator architecture mushroom applied cryptology. Devadas established the 2015 ACM/IEEE Richard Newton present, the 2017 IEEE W. Wallace McDowell award instruction the 2018 IEEE River A. Desoer award stick up for his investigation in hearty hardware. Pacify is a Fellow hill the ACM and IEEE. He psychiatry a MacVicar Faculty Gentleman, an Everett Moore Baker and a Bose grant recipient, thoughtful MIT's upper teaching honors.
Abstract:
As representation world becomes more conterminous, privacy enquiry becoming harder to free from blame. From popular media services to Net service providers to state-sponsored mass-surveillance programs, many outlets collect discerning information travel the patrons and depiction communication among them – often left out the ultimate consumers ever secret about strike. In take on, many Information superhighway users plot turned denomination end-to-end coding, like Alarm and TLS, to safeguard the content of rendering communication. Sadly, these complex do approximately to leather the metadata of interpretation communication, much as when and tackle whom a user deference communicating. Person of little consequence scenarios where the metadata are reactive,
•
Paper 2021/687
Towards Understanding Practical Randomness Beyond Noise: Differential Privacy and Mixup
Hanshen Xiao and Srinivas Devadas
Abstract
Information-theoretical privacy relies on randomness. Representatively, Differential Privacy (DP) has emerged as the gold standard to quantify the individual privacy preservation provided by given randomness. However, almost all randomness in existing differentially private optimization and learning algorithms is restricted to noise perturbation. In this paper, we set out to provide a privacy analysis framework to understand the privacy guarantee produced by other randomness commonly used in optimization and learning algorithms (e.g., parameter randomness). We take mixup: a random linear aggregation of inputs, as a concrete example. Our contributions are twofold. First, we develop a rigorous analysis on the privacy amplification provided by mixup either on samples or updates, where we find the hybrid structure of mixup and the Laplace Mechanism produces a new type of DP guarantee lying between Pure DP and Approximate DP. Such an average-case privacy amplification can produce tighter composition bounds. Second, both empirically and theoretically, we show that proper mixup comes almost free of utility compromise.
•
Speakers
Edward Suh
NVIDIA
Bio: G. Edward Suh is a Senior Director of Research, and leads a group in security and privacy research. He is also an Adjunct Professor in the School of Electrical and Computer Engineering at Cornell University, where he served on the faculty from 2007 to 2023. Before joining NVIDIA, he was a Research Scientist in the Fundamental AI Research (FAIR) team at Meta. He earned a B.S. in Electrical Engineering from Seoul National University and an M.S. and a Ph.D. in Electrical Engineering and Computer Science from the Massachusetts Institute of Technology (MIT). His research interests include computer systems in general with particular focus on computer architecture and security. His recent research focuses on building secure computing systems for secure and private AI, and using AI to improve the security of computer systems. His past research received multiple test-of-time awards and is widely recognized for the impact at the intersection of hardware and security. For example, his work on Physical Unclonable Function (PUF) is now used in commercial products such as Xilinx FPGAs for storing secret keys. His work on the AEGIS secure processor received a test-of-time award for its contribution for trusted execution environments deployed across