A Survey of Recent Attacks and Mitigation on FPGA Systems
Shijin Duan, Wenhao Wang, Yukui Luo, Xiaolin Xu
Abstract
The emergence of a large variety of compute-intensive applications has made hardware accelerators a new necessity to deploy the corresponding high-complexity algorithms, such as the Deep Neural Network (DNN). Thanks to the flexibility from hardware reconfiguration and high power efficiency, field-programmable gate array (FPGA) has been widely utilized for building DNN hardware accelerators. In particular, FPGA has become one of the most popular edge platforms for deep-learning algorithm acceleration and machine learning as a service (MLaaS) in the cloud. Although significantly improving the performance of DNN algorithms, these FPGA-based accelerators also face unique and novel security vulnerabilities that the community should pay more attention to. This paper systematically reviews the state-of-the-art research on FPGA-based hardware acceleration systems and their security issues, discusses the feasibility of existing defense solutions, and envisions future research directions.