Authors: |
- Jingdian Ming , Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China; School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China
- Huizhong Li , Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China; School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China
- Yongbin Zhou , Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China; School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China; School of Cyber Security, Nanjing University of Science and Technology, Nanjing, China
- Wei Cheng , Télécom Paris, Polytechnique de Paris, Palaiseau, France
- Zehua Qiao , Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China; School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China
|
Abstract: |
Addition chain is a well-known approach for implementing higher-order masked SBoxes. However, this approach induces more computations of intermediate monomials over F2n, which in turn leak more information related to the sensitive variables and may decrease its side-channel resistance consequently. In this paper, we introduce a new notion named polygon degree to measure the resistance of monomial computations. With the help of this notion, we select several typical addition chain implementations with the strongest or the weakest resistance. In practical experiments based on an ARM Cortex-M4 architecture, we collect power and electromagnetic traces in consideration of different noise levels. The results show that the resistance of the weakest masked SBox implementation is close to that of an unprotected implementation, while the strongest one can also be broken with fewer than 1,500 traces due to extra leakages. Moreover, we study the resistance of addition chain implementations against profiled attacks. We find that some monomials with smaller output size leak more information than the SBox output. The work by Duc et al. at JOC 2019 showed that for a balanced function, the smaller the output size is, the less information is leaked. Thus, our attacks demonstrate that this property of balanced functions does not apply to unbalanced functions. |