Security Threats in Approximate Computing Systems
Proceedings of the 2019 on Great Lakes Symposium on VLSI(2019)
摘要
Approximate computing systems improve energy efficiency and computation speed at the cost of reduced accuracy on system outputs. Existing efforts mainly explore the feasible approximation mechanisms and their implementation methods. There is limited work that investigates the security threats brought by approximate computing. To fill this gap, we first analyze the approximate mechanisms used in approximate system, software, storage, and arithmetic circuits, and then propose potential attacks that will challenge the integrity and security of approximate systems. Some illustrative examples are provided accordingly to showcase the consequences of the proposed new attacks.
更多查看译文
关键词
approximate computing, dram, error resilience, hardware security, image processing, machine learning, pcm, sram
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络