부채널 분석을 이용한 DNN 기반 MNIST 분류기 가중치 복구 공격 및 대응책 구현
Weight Recovery Attacks for DNN-Based MNIST Classifier Using Side Channel Analysis and Implementation of Countermeasures
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
딥러닝 기술은 자율 주행 자동차, 이미지 생성, 가상 음성 구현 등 다양한 분야에서 활용되고 있으며 하드웨어 장치에서 고속 동작을 위해 딥러닝 가속기가 등장하게 되었다. 그러나 최근에는 딥러닝 가속기에서 발생하는 부채널 정보를 이용한 내부 비밀 정보를 복구하는 공격이 연구되고 있다. 본 논문에서는 DNN(Deep Neural Network) 기반 MNIST 숫자 분류기를 마이크로 컨트롤러에서 구현한 후 상관 전력 분석(Correlation Power Analysis) 공격을 시도하여 딥러닝 가속기의 가중치(weight)를 충분히 복구할 수 있음을 확인하였다. 또한, 이러한 전력 분석 공격에 대응하기 위해 전력 측정 시점의 정렬 혼돈(misalignment) 원리를 적용한 Node-CUT 셔플링 방법을 제안하였다. 제안하는 대응책은 부채널 공격을 효과적으로 방어할 수 있으며, Fisher-Yates 셔플링 기법을 사용하는 것보다 추가 계산량이 1/3보다 더 줄어듦을 실험을 통해 확인하였다.
Abstract
Deep learning technology is used in various fields such as self-driving cars, image creation, and virtual voice implementation, and deep learning accelerators have been developed for high-speed operation in hardware devices. However, several side channel attacks that recover secret information inside the accelerator using side-channel information generated when the deep learning accelerator operates have been recently researched. In this paper, we implemented a DNN(Deep Neural Network)-based MNIST digit classifier on a microprocessor and attempted a correlation power analysis attack to confirm that the weights of deep learning accelerator could be sufficiently recovered. In addition, to counter these power analysis attacks, we proposed a Node-CUT shuffling method that applies the principle of misalignment at the time of power measurement. It was confirmed through experiments that the proposed countermeasure can effectively defend against side-channel attacks, and that the additional calculation amount is reduced by more than 1/3 compared to using the Fisher-Yates shuffling method.
참고문헌 (12)
- L. Batina, "CSI NN : Reverse engineering of neural network architectures through Electromagnetic side channel", Proceedings of the 28th USENIX Conference on Security Symposium, pp.515~532 (2019)
- 박수진, "부채널 분석을 이용한 딥러닝 네트워크 신규 내부 비밀정보 복원 방법 연구", 정보보호학회논문지, 32(5), pp.855~867 (2022), 한국정보보호학회[DOI]
- Y. Liu, "Mitigating reverse engineering attacks on deep neural networks", 2019 IEEE Computer Society Annual Symposium on VLSI(ISVLSI), pp.657~662 (2019)
- Q. Fang, "Voltage scaling-agnostic counteraction of side-channel neural net reverse engineering via machine learning compensation and multi-level shuffling", 2023 IEEE Symposium on VLSI Technology and Circuits, pp.1~2 (2023)
- K. Athanasiou, "Masking feedforward neural networks against power analysis attacks", proceedings on privacy enhancing technologies, pp.501~521 (2022)
- Y. LeCun, "Deep learning", Nature, 521, pp.436~444 (2015)
- S. Mangard, "Simple power analysis-power analysis attacks revealing the secrets of smart cards", pp.101~118 (2008), Springer-Verlag
- P. Kocher, "Differential power analysis", Advances in Cryptology, CRYPTO’ 99, LNCS, 1666, pp.388~397 (1999)
- E. Brier, "Correlation power analysis with a leakage model", CHES’04, LNCS, 3156, pp.16~29 (2004)
- A. G. Bayrak, "An architecture-independent instruction shuffler to protect against side-channel attacks", ACM Transactions on Architecture and Code Optimization(TACO), pp.1~19 (2012)
- N. Veyrat-Charvillon, "Shuffling against side-channel attacks : a comprehensive study with cautionary note", Advances in Cryptology–ASIACRYPT, pp.740~757 (2012)
- R. A. Fisher, "Statistical table for bio, agriculture, and medical research" (1963), Oliver & Boyd