Abstract: Fueled by the proliferation of sensing data and the advancement of machine learning (ML), intelligence is becoming a household brand in many Cyber-Physical Systems (CPS) and Internet-of-Things (IoT) applications. While enticing, anchoring intelligent data processing in efficiency and trustworthiness under the stringent requirements for end-to-end latency (often real time), energy, and resource, is often challenging but crucial in such applications. In this talk, I will present my vision on rooting intelligence in constrained CPS and IoT applications from several radically different perspectives on computing, automation, and security. The talk starts with the rethinking of data compression design to ease the communication latency bottlenecking the edge-cloud collaborative inference, followed by algorithm-hardware co-design for accelerating and automating multi-model on-device inference with the latency and accuracy adaptive to the environment changes for energy harvesting IoT devices. Then I will briefly discuss detection and recovery solutions for ensuring safe ML inference on hardware accelerators against execution disturbances (e.g. physical fault attacks) with guarantees of real-time, low-cost, and high-performance. Our prospects on the research in this direction will be also given at the end of this talk, offering the audiences an alternative thinking about efficiency and security by design tailored to modern intelligent CPS and IoT.

Short Bio: Wujie Wen is an assistant professor in the Department of Electrical and Computer Engineering (ECE) at Lehigh University. He received his Ph.D. from University of Pittsburgh in 2015. He earned his B.S. and M.S. degrees from Beijing Jiaotong University and Tsinghua University, Beijing, China, in 2006 and 2010, respectively. He was an assistant professor in the ECE department of Florida International University, Miami, FL, during 2015-2019. His research interests include efficient, reliable, and secure computing, design automation, as well as their applications to embedded, IoT and Cyber-Physical Systems. His works have been published widely across venues in design automation, security, machine learning/AI etc., including DAC, ICCAD, DATE, HPCA, ICPP, USENIX Security, ACSAC, HOST, NeurIPS, CVPR, ECCV, AAAI etc. He received best paper nominations from all 4 major electronic design automation conferences-DAC, ICCAD, DATE and ASP-DAC. Dr Wen served as the General Chair and Technical Program Chair of ISVLSI 2018 and 2019, respectively, as well as program committee for many conferences such as DAC, ICCAD, DATE, HPCA etc. He is an associated editor of Neurocomputing and IEEE Circuit and Systems (CAS) Magazine. His research projects are sponsored by NSF, AFRL and Florida Center for Cybersecurity etc.