TY - GEN
T1 - A comparison of weight initializers in deep learning-based side-channel analysis
AU - Li, Huimin
AU - Krček, Marina
AU - Perin, Guilherme
PY - 2020
Y1 - 2020
N2 - The usage of deep learning in profiled side-channel analysis requires a careful selection of neural network hyperparameters. In recent publications, different network architectures have been presented as efficient profiled methods against protected AES implementations. Indeed, completely different convolutional neural network models have presented similar performance against public side-channel traces databases. In this work, we analyze how weight initializers’ choice influences deep neural networks’ performance in the profiled side-channel analysis. Our results show that different weight initializers provide radically different behavior. We observe that even high-performing initializers can reach significantly different performance when conducting multiple training phases. Finally, we found that this hyperparameter is more dependent on the choice of dataset than other, commonly examined, hyperparameters. When evaluating the connections with other hyperparameters, the biggest connection is observed with activation functions.
AB - The usage of deep learning in profiled side-channel analysis requires a careful selection of neural network hyperparameters. In recent publications, different network architectures have been presented as efficient profiled methods against protected AES implementations. Indeed, completely different convolutional neural network models have presented similar performance against public side-channel traces databases. In this work, we analyze how weight initializers’ choice influences deep neural networks’ performance in the profiled side-channel analysis. Our results show that different weight initializers provide radically different behavior. We observe that even high-performing initializers can reach significantly different performance when conducting multiple training phases. Finally, we found that this hyperparameter is more dependent on the choice of dataset than other, commonly examined, hyperparameters. When evaluating the connections with other hyperparameters, the biggest connection is observed with activation functions.
KW - Deep learning
KW - Side-channel analysis
KW - Weight initialization
UR - http://www.scopus.com/inward/record.url?scp=85094111177&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-61638-0_8
DO - 10.1007/978-3-030-61638-0_8
M3 - Conference contribution
AN - SCOPUS:85094111177
SN - 9783030616373
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 126
EP - 143
BT - Applied Cryptography and Network Security Workshops - ACNS 2020 Satellite Workshops, AIBlock, AIHWS, AIoTS, Cloud S and P, SCI, SecMT, and SiMLA, Proceedings
A2 - Zhou, Jianying
A2 - Ahmed, Chuadhry Mujeeb
A2 - Conti, Mauro
A2 - Losiouk, Eleonora
A2 - Au, Man Ho
A2 - Batina, Lejla
A2 - Li, Zhou
A2 - Lin, Jingqiang
A2 - Luo, Bo
A2 - Majumdar, Suryadipta
A2 - Meng, Weizhi
A2 - Ochoa, Martín
A2 - Picek, Stjepan
A2 - Portokalidis, Georgios
A2 - Wang, Cong
A2 - Zhang, Kehuan
PB - Springer
T2 - 2nd ACNS Workshop on Application Intelligence and Blockchain Security, AIBlock 2020, 1st ACNS Workshop on Artificial Intelligence in Hardware Security, AIHWS 2020, 2nd ACNS Workshop on Artificial Intelligence and Industrial IoT Security, AIoTS 2020, 2nd ACNS Workshop on Cloud Security and Privacy, Cloud S and P 2020, 1st ACNS Workshop on Secure Cryptographic Implementation, SCI 2020, 1st ACNS Workshop on Security in Mobile Technologies, SecMT 2020, and 2nd ACNS Workshop on Security in Machine Learning and its Applications, SiMLA 2020, held in parallel with the 18th International Conference on Applied Cryptography and Network Security, ACNS 2020
Y2 - 19 October 2020 through 22 October 2020
ER -