Quantitative Analysis of Information Leakage Through Hardware RSA Timing Channel Based on Entropy Theory
Bao-Lei MAO,Wei HU,De-Jun MU,Hui-Xiang ZHANG,Yu TAI,Liang HONG
DOI: https://doi.org/10.11897/SP.J.1016.2018.00426
2018-01-01
Chinese Journal of Computers
Abstract:The RSA algorithm is a widely deployed public key cipher for data encryption and digital signature,whose security has drawn attention from both academic and industry fields.Its security relies on both the computation complexity of breaking the algorithm itself and the security of the encryption key.Generally,it is much easier to recover the encryption key than break the RSA algorithm through power and timing side channel analysis.Previous work primarily focuses on timing side channels in software RSA implementations,without in-depth studying the effect of hardware architecture on timing channel security.Although there is work for detecting and isolating timing channel based on formal verification of timing information flow or building timing channel free hardware design by incorporating new type system into the hardware design language,they can only provide qualitative analysis of timing channel,lacking effective model to perform quantitative analysis of hardware timing channel security.In this work,we will concentrate on hardware RSA cores and provide a quantitative analysis model to evaluate such timing channel leakage.Firstly,we introduce hardware RSA timing channel and its threat model.We then employ the entropy theory to set up timing attack model and quantitative analysis model for RSA architecture timing channel.Besides,we attack RSA implementation based on entropy and variance analysis,respectively.In order to demonstrate the effectiveness of entropy in quantifying hardware RSA implementation leakage,we perform quantitative analysis of different key-pairs information leakage within the same RSA architecture,quantify information leakage for different RSA architecture implementations with timing channel algorithm optimization techniques,evaluate the effect of timing channel countermeasure on reducing information leakage;and also attack each RSA implementation to validate the effectiveness of our quantitative analysis model.Finally,we evaluate the effect of different algorithm optimizations,timing channel mitigation techniques and countermeasures on design complexity in terms of timing channel,performance and resource utilization.Experimental results show that entropy metric can be used to attack RSA timing channel and it can increase the success rate by combining variance analysis with entropy analysis.Entropy metric can be used to quantitatively analyze information leakage from timing channel in RSA hardware architectures effectively and efficiently,which provides an effective theoretical basis and test methodology to assess the severity of timing channel information leakage.In addition,entropy metric can help designers to tradeoff security requirements and design overheads such as performance and resource utilization,which provides a potential security metric to integrate timing channel security with traditional design metrics (e.g.area and performance) to characterize the hardware in more detail.