Reliable and Secure Edge Computing for Data and Compute-intensive Applications: Challenges and directions

ABSTRACT

Modern edge computing paradigm finds itself to be increasingly driven by the need to support data- and compute-intensive applications, that are often ML/AI-based and are generated by a variety of critical use-cases ranging from public safety and emergency response to cross-campus bioinformatics and elder-care. Often, such applications vary greatly in terms of their performance, reliability, security, and privacy requirements, nature of underlying algorithms, data size and type, and hardware heterogeneity. Simultaneously, many of the underlying use-cases operate with systems and within environments that are constrained and dynamic in terms of resource availability and system disruptions. These often-diverging factors create challenging inter-conflict problems among system design objectives that are simultaneously data-, application-, system-, and environment-driven. Unfortunately, the traditional distributed resource management strategies are unsuited to address such inter-conflicts. This talk presents the unique challenges of edge resource management for a few of such exemplary applications: i) rapid 3D reconstruction of dynamic scenes, ii) energy-efficient DNN inference, iii) trustworthy volunteer edge-cloud management, and iv) secure and privacy-preserving ML for collaborative edge. The talk highlights our initial approaches to design novel and intelligent models and algorithms to strike careful balance among such diverging design objectives. In particular, the talk discusses algorithms, architectures, and early results from three NSF and Google funded projects.

BIOGRAPHY

Saptarshi Debroy is an Associate Professor of Computer Science at Hunter College and a member of the doctoral faculties at The Graduate Center of The City University of New York (CUNY). He received his PhD degree from University of Central Florida in 2014. Before joining academia, he worked as research scientist for Motorola Inc. and Blue Coat Systems. His current research interest is reliable and secure Edge AI. He has published more than 50 papers in top-tier peer-reviewed journals and conference proceedings. He currently manages and previously managed multiple research and education projects funded by federal agencies, such as National Science Foundation (NSF), Department of Energy (DOE) and by corporations, such as Google Inc. and Verizon Foundation. Some of his accomplishments include NSF Early career award, nomination for Presidential Early Career Awards for Scientists and Engineers (PECASE), best paper awards, and university gold medal. He is a senior member of ACM and IEEE