ECE Seminars

UPCOMING EVENTS- Spring 2020: 

Thursday, February 13, 2020
4:25pm 
Packard Lab 466
Speaker: Tamzidul Hoque  
Trust, but Tolerate and Verify: Ecosystem of Technologies for Trusted Microelectronics

Abstract: 

In the new era of intelligent connectivity, electronic devices have gained unprecedented access to our daily lives. However, the economic advantage of outsourcing the design, fabrication, and assembly process of electronic components has driven the semiconductor industry to rely on various foreign vendors across the globe. These untrusted vendors can introduce stealthy, malicious modifications to the hardware design, also known as hardware Trojans, to cause a functional failure or leak secret information during field operation. Even after deployment, the hardware design is vulnerable to various invasive in-field attacks by malicious end-users that can disrupt trusted execution.
In this talk, I will present an ecosystem of technologies that can verify the trustworthiness of the hardware designs before deployment and integrate the ability to tolerate various attacks in the field. First, I will outline the security and trust issues in the integrated circuit (IC) lifecycle. Next, I will present two solutions to verify the trustworthiness of ICs that can identify hardware Trojans before deployment, in both pre-silicon and post-silicon stages of the supply chain. Similarly, for post-deployment phases, I will present two solutions. The first one enables Trojan-tolerant computing in the field that can be used as the last line of defense for applications where trust verification is not feasible. The next technology in the ecosystem is developed to protect reconfigurable hardware that is vulnerable to tampering of its bitstream in the field. I will illustrate how a machine learning guided hardware obfuscation method can enable tamper-tolerance in reconfigurable hardware. In conclusion, I will discuss the future security challenges that I intend to address to establish hardware as the trust anchor for emerging applications.

Bio: 

Tamzidul Hoque is a Ph.D. candidate in the Electrical and Computer Engineering at the University of Florida (UF). His Ph.D. research is focused on hardware security, with emphasis on hardware Trojan detection and hardware intellectual property protection from piracy, reverse engineering, and tampering. He has worked within the Advanced Security Research and Government Group (ASRG) in Cisco as a Ph.D. intern during the summer 2016 and 2017. In Cisco, Tamzidul was selected to work with their Trustworthy Technologies (TT) group that is assigned to secure the entire product line. Tamzidul’s research in the area of security has so far resulted in more than 20 peer-reviewed publications in premier journals and international conferences, including lead-authored articles in IEEE International Test Conference, IEEE Consumer Electronics Magazine, and IEEE Design & Test. As a recognition of his research contribution, Tamzidul received the ECE Graduate Research Excellence Award from UF in 2018. The technical demonstration of his work has received five awards in leading hardware security conferences, including IEEE Hardware Oriented Security and Trust.

 
Tuesday, February 11, 2020
4:25pm 
Packard Lab 466
Speaker: Dr. Marco Donato   
DNN-Based Edge Computing via Full-Stack Co-Design Approach

Abstract: 

Augmented/Virtual Reality platforms, Real-Time Translators, and Intelligent Personal Assistants are emerging applications within our grasp thanks to continuous developments in Deep Neural Network (DNN) models. DNN hardware accelerators are paving the way for such applications to be deployed to embedded devices. However, data movement remains a critical bottleneck and CMOS memory technologies struggle to keep up with the increase in memory footprint for state-of-the-art DNN models, forcing the system to perform costly off-chip DRAM memory access. As the future of computing is shaped by the way we store and process large amounts of information, a need for design solutions at the intersection of devices, circuits, architectures, and applications emerges. This talk will present full-stack design methodologies that leverage embedded non-volatile memories (eNVMs) as a dense, approximate storage solution to reduce DRAM accesses by storing DNN models entirely on chip. In evaluating the implications of building eNVM-based computing systems, it is critical to take into account the non-idealities of many eNVM implementations. Multi-level cell storage offers the opportunity for higher capacity, but introduces reliability issues. Moreover, the high energy and latency costs of eNVM writes, together with limited memory endurance, set a limit on how frequently and efficiently the memory content can be updated. I will show that all these limitations can be circumvented by adopting device, architectural, and algorithmic co-design optimizations, making eNVMs a viable solution for energy-efficient DNN inference at the edge.

Bio: 

Dr. Marco Donato is a Research Associate in the John A. Paulson School of Engineering and Applied Sciences at Harvard University. His research focuses on the design of novel embedded memory subsystems and circuitry in advanced CMOS technology nodes with applications to machine learning hardware accelerator SoCs. Before joining Harvard, Dr. Donato received his Ph.D. from Brown University where he worked on developing automated tools for noise-tolerant circuit architecture design, and physically accurate and computationally efficient simulation techniques for evaluating the effect of thermal and random telegraph signal (RTS) noise in nanoscale sub-threshold CMOS circuits. Dr. Donato holds a Master and Bachelor’s degree from the University La Sapienza in Rome.

 


PAST EVENTS- Fall 2019:

Monday, November 11, 2019
4:15pm 
Packard Lab 466

Speaker: Dr. Peng Li  

Learning Mechanisms and Hardware Design for Computation with Spiking Neurons

Abstract: 

As one form of brain-inspired computing, spiking neural networks (SNN) have recently gained momentum. This is fueled by in part by advancements in emerging devices and neuromorphic hardware, e.g., availability of Intel Loihi and IBM TrueNorth neuromorphic chips, promising ultra-low energy event-driven processing of large amounts of data. Nevertheless, major challenges are yet to be conquered to make spikebased computation a competitive choice for real-world applications. This talk will present a multi-faceted SNN research approach: 1) empowering SNNs by exploring computationally-powerful feedforward and recurrent architectures; 2) tackling major challenges in training complex SNNs by developing biologically plausible learning mechanisms and error backpropagation operating on top of spiking discontinuities; and 3) enabling efficient FPGA spiking neural processors with integrated on-chip learning via algorithm-hardware co-optimization.

Bio: 

Peng Li received the Ph. D. degree from Carnegie Mellon University in 2003. He was on the faculty of Texas A&M University from August 2004 to June 2019. Since July 2019, he has been with the University of California at Santa Barbara as a professor of Electrical and Computer Engineering.His research interests are in integrated circuits and systems, electronic design automation, brain-inspired computing, and computational brain modeling. Li’s work has been recognized by an ICCAD Ten Year Retrospective Most Influential Paper Award, four IEEE/ACM Design Automation Conference (DAC) Best Paper Awards, an Honorary Mention Best Paper Award from ISCAS, an IEEE/ACM William J. McCalla ICCAD Best Paper Award, two SRC Inventor Recognition Awards, two MARCO Inventor Recognition Awards, and an NSF CAREER Award. He was honored by the ECE Outstanding Professor Award, and was named a TEES Fellow, a William O. and Montine P. Head Faculty Fellow, and a Eugene Webb Fellow by the College of Engineering at Texas A&M University. He was the Vice President for Technical Activities of the IEEE Council on Electronic Design Automation from Jan. 2016 to Dec. 2017. He is a Fellow of the IEEE and has consulted for Intel and several Silicon Valley startup companies.


 

Wednesday, October 2, 2019

4:15pm 
Packard Lab 466
Universal Features- Information Extraction and Data-Knowledge Integration 

Speaker: Dr. Lizhong Zheng, MIT  

Abstract:

With the growing demand of using data analytics in a wide range of applications, a key research challenge has emerged to represent data in a generic semantic space, where we need to have a quantitative way to represent the useful information and knowledge succinctly, and at an abstract level. The key issues include how to define a universal interface for knowledge representation, how to manage and integrate the knowledge from multiple data sources, how to utilize domain knowledge, and how to cope with non-ideal situations such as the disparity in the quality of different datasets and precision losses in the processing. There are numerous algorithms as possible ways to achieve such goals. Particularly, neural networks are expected to play a key role. The main difficulty is that we still do not have a complete theory about deep learning, to identify exactly what knowledge is learned by neural networks, what hidden assumptions are needed for the desirable performance. In this talk, we try to address this problem by developing a theoretical structure to measure the meaning of information by its relevance to specific inference problems, and from that we explain the behavior of neural networks as extracting “universal features”, defined as the solution to a specific optimization problem. This helps us not only to understand the learning process inside a large neural network, but also to draw connections to a number of well-known concepts in statistics and other learning algorithms. Based on this theoretic framework, our goal is to develop more flexible, robust, and interpretable data embedding algorithms.

Bio: 

Lizhong Zheng received the B.S and M.S. degrees, in 1994 and 1997 respectively, from the Department of Electronic Engineering, Tsinghua University, China, and the Ph.D. degree, in 2002, from the Department of Electrical Engineering and Computer Sciences, University of California, Berkeley. Since 2002, he has been working at MIT, where he is currently a professor of Electrical Engineering. His research interests include information theory, statistical inference, communications, and networks theory. He received Eli Jury award from UC Berkeley in 2002, IEEE Information Theory Society Paper Award in 2003, and NSF CAREER award in 2004, and the AFOSR Young Investigator Award in 2007. He served as an associate editor for IEEE Transactions on Information Theory, and the general co-chair for the IEEE International Symposium on Information Theory in 2012. He is an IEEE fellow.


 

Monday, September 23, 2019
1:30pm 
Steps 101
Co-Sponsored with the Institute for Cyber Physical Infrastructure & Energy
Information for Control and Sensing in Cyber Physical Systems: From Maxwell's Demon to Millimeter Wave

Speaker: Dr. Husheng Li, University of Tennessee  

Abstract: 

The talk is focused on the efficiency and gleaning of information in cyber physical systems (CPSs). For evaluating the interdependence of communications and controls in CPSs, we study the control of entropy (or equivalently uncertainty) via communications in CPSs. We consider the controller as the Maxwell's demon that decimates the system entropy. Due to the second law of thermodynamics, the system entropy cannot be spontaneously decreased. Therefore, to reduce the system entropy, the controller needs external information communicated from sensors. The information efficiency of reducing entropy is studied for both finite and continuous state systems. Then, we will discuss how to leverage the millimeter wave communications in 5G systems for the purpose of sensing, as a `bonus' of wireless communications. Detection, tracking and imaging for objects in the millimeter wave band will be discussed, based on the millimeter wave signals blocked or reflected by the objects of interest. Both experimental and numerical results on this `bonus' sensing mechanism will be introduced.

Bio:

Husheng Li received the BS and MS degrees in electronic engineering from Tsinghua University, Beijing, China, in 1998 and 2000, respectively, and the Ph.D. degree in electrical engineering from Princeton University, Princeton, NJ, in 2005. From 2005 to 2007, he worked as a senior engineer at Qualcomm Inc., San Diego, CA. In 2007, he joined the EECS department of the University of Tennessee, Knoxville, TN, as an assistant professor, and has been promoted to full professor. His research is mainly focused on statistical signal processing, wireless communications, networking, smart grid and game theory. Dr. Li is the recipient of the Best Paper Awards of EURASIP Journal of Wireless Communications and Networks, 2005, EURASIP Journal of Advances in Signal Processing, 2015, IEEE Globecome 2017, IEEE ICC 2011 and IEEE SmartGridComm 2012, and the Best Demo Award of IEEE Globecom, 2010.