Webinar: Hardware Based Attacks and Solutions on Proprietary Deep Learning Models

Speaker:

Dr. Ankur Srivastava, Director of Institute for Systems Research at University of Maryland

Abstract:

Deep learning (DL) algorithms are extensively used for analyzing big data in several domains including image classification, natural language processing, autonomous transportation, smart health, financial management, social networks, etc.. The key factors attributed to the unprecedented success of these algorithms are (i) availability of a massive and mostly labeled training dataset, (ii) allocation of powerful computing resources as well as vast amounts of network training time, and also (iii) substantial domain expertise of DL model developers to obtain highly accurate models. Therefore, well trained DL models are considered to be intellectual property (IP) of the owner as significant cost is incurred behind their training process to gain a competitive edge in business.

Speaker Bio:

Ankur Srivastava, the seventh director of the Institute for Systems Research, has a joint appointment in the Electrical and Computer Engineering Department and ISR. Dr. Srivastava received his B.Tech in Electrical Engineering from Indian Institute of Technology Delhi in 1998 and PhD in Computer Science from UCLA in 2002. He was awarded the prestigious Outstanding Dissertation Award from the CS department of UCLA in 2002. His primary research interests lie in the field of high performance, low power and secure electronic systems and applications such as computer vision, data and storage centers and sensor networks. He has published numerous papers on these topics at prestigious venues. He has been a part of the technical program & organizing committees of several conferences such as ICCAD, DAC, ISPD, ICCD, GLSVLSI, HOST and others. He has served as the associate editor for IEEE Transactions on VLSI, IEEE Transactions on CAD and INTEGRATION: VLSI Journal. His research and teaching contributions have also been recognized through various awards.