Special Sessions

Special sessions will be held throughout Monday, September 30, 2024 and Wednesday, October 2, 2024.

SS 1. Detecting and Defending Vulnerabilities in Heterogeneous and Monolithic Systems: Current Strategies and Future Directions

The advancements in VLSI technology, computer architecture, and fabrication methodologies together have led to the emergence of increasingly sophisticated and complex system-on-chips (SoCs) and processors. In addition to offering a wide range of performance enhancements, these advancements also expose various exploits and vulnerabilities that emerge from both hardware design and software functionality, along with their interactions, that have never been experienced in the past. Detecting these vulnerabilities and functional exploits (bugs) in the early phase is crucial for rectifying the exploits, preserving the system integrity and mitigating the post-production re-engineering costs.

  1. Towards a Robust Metrology for Heterogeneous System-on-Chip Security; Guru Prasadh Venkataramani (George Washington University)
  2. Secure Embedded Systems’ Design by Leveraging Hardware-Software Limitation and Interactions; Sai Manoj Pudukotai Dinakarrao (George Mason University)
  3. Snowflake IoT: Ultra-Low-Cost Diversity Defenses; Todd Austin (University of Michigan)
  4. Securing Large Monolithic Systems: Challenges and Opportunities; Ashish Venkat (University of Virginia)

SS 2. Emerging Architecture Design, Control, and Security Challenges in Software Defined Vehicles (SDVs)

Software Defined Vehicles (SDVs) represent a paradigm shift in the automotive industry, where vehicles are increasingly controlled and managed through software, while relying less on mechanical and hardware components. While this allows considerable flexibility in the introduction of new “smart” features and fast tracks innovations in multiple domains, it also creates new challenges and opportunities in architecture design, control, and security. By adopting modular architectures, adaptive control strategies, and robust security measures, SDVs can pave the way for a safer and more efficient future of transportation. In this session, we aim to cover perspectives from both, industry and academia, in this area. Our goal is to provide embedded systems researchers a snapshot of recent developments and emerging challenges in SDV from the perspective of architecture design, control, and security.

  1. Emerging In-Vehicle Architectures in the Age of SDV; Khaja Shazzad (General Motors R&D)
  2. New Paradigms for Automotive Control in the Era of SDV; Samarjit Chakraborty (University of North Carolina at Chapel Hill)
  3. Emerging Robustness: Robust Perception with Embedded Systems in SDVs; Sudeep Pasricha (Colorado State University)
  4. Emerging Security: Covert Channel and In-vehicle Network Spoofing Attacks on Embedded Systems in SDVs; Amit Kumar Singh (University of Essex)

SS 3. Design for Environmental Sustainability in Computing

This session is dedicated to spearheading advancements in the realm of design for environmental sustainability in computing. It is a crucible for innovation, providing a unique platform for thought leaders, and researchers to address the complex challenges and opportunities presented by sustainable computing.

  1. Modeling and Optimizing the Carbon Objective for Sustainable Architectures and Systems; Alex K. Jones (University of Pittsburgh)
  2. Sustainable Deployment of Deep Neural Networks on Non-Volatile Compute-in-Memory Accelerators; Yiyu Shi (University of Notre Dame)
  3. End-To-End Carbon Footprint Assessment and Modeling of Deep Learning; Fan Chen (Indiana University Bloomington)

SS 4. Estimation and Optimization of DNNs for Embedded Platforms

Efficient implementation of a Deep Neural Network (DNN) on a given platform under tight constraints is challenging due to many non-linear dependencies. Small changes in the DNN or the platform configuration often have disproportional effects on the performance. This special session will explore state-of-the art methods for optimization and estimation of DNNs on tightly constrained embedded platforms.

  1. TiVisual Language Models for Edge AI 2.0; Song Han (MIT)
  2. DNN Model Optimization and Implementation for Embedded Systems; Lin Meng (Ritsumeikan University)
  3. Latency estimation; Axel Jantsch (TU Wien)
  4. Multi-level Performance Estimation of Multi-instance AI Compute Platforms; Oliver Bringmann (University of Tübingen)

SS 5. Neuro-Symbolic Architecture Meets Large Language Models: A Memory-Centric Perspective

This session calls for an in-depth exploration of the integration of neuro-symbolic architectures (NeSy) with large language models (LLMs) in the context of edge computing. We seek to address critical questions related to the fusion of symbolic reasoning and neural networks, focusing on the theoretical foundations, computational challenges, and advancements associated with harnessing memory-centric processing. Through this special session, we aim to bring together experts and researchers to delve into the intersection of these cutting-edge technologies, fostering discussions on topics ranging from memory-aware quantization of LLMs and in-memory symbolic reasoning to ethical and privacy considerations in distributed edge intelligence.

  1. Biologically Inspired Computing Architectures and Circuits for Embodied Intelligence; Arijit Raychowdhury (Georgia Institute of Technology)
  2. Efficient LLMs: Innovations in Quantization, Memory, and Attention; Priyadarshini Panda (Yale University)
  3. Energy-Efficient Computational Memories as Heterogenous CMOS+X Platforms for Neuro-Symbolic and Transformer Architectures; Haitong Li (Purdue University)
  4. Von Neumann-like Architecture for Computing with High-Dimensional Vectors Motivated by Neuroscience and Psychology; Pentti Kanerva (University of California at Berkeley)
  5. The Application of Spiking Neural Networks (SNNs) in Large Language Models and Their Implementation with Compute-in-Memory (CIM); Yiran Chen (Duke University)