Kazukuni Kobara

In the field of machine learning security, as one of the attack surfaces especially for edge devices, the application of side-channel analysis such as correlation power/electromagnetic analysis (CPA/CEMA) is expanding. Aiming to evaluate the leakage resistance of neural network (NN) model parameters, i.e. weights and biases, we conducted a feasibility study of CPA/CEMA on floating-point (FP) operations, which are the basic operations of NNs. This paper proposes approaches to recover weights and biases using CPA/CEMA on multiplication and addition operations, respectively. It is essential to take into account the characteristics of the IEEE 754 representation in order to realize the recovery with high precision and efficiency. We show that CPA/CEMA on FP operations requires different approaches than traditional CPA/CEMA on cryptographic implementations such as the AES.

As an alternative to tamper-resistant modules, PUFs (Physically Unclonable Functions) [1], [6], [7], [8], [9], [10], [11], [14], [15], [18], [25], [26] have received much attention due to their unclonable physical characteristics. In this paper, we propose key establishment (KE1 and KE2) protocols using PUFs which can mitigate side-channel attacks and satisfy two security requirements (i.e., Freshness and Non-Regeneration). Also, we discuss security of the KE1 and KE2 protocols, and compare with the previous PUF-based key exchange protocols [3], [4], [27].

A PAKR (Password-Authenticated Key Retrieval) protocol and its multi-server system allow one party (say, client), who has a rememberable password, to retrieve a long-term static key in an exchange of messages with at least one other party (say, server) that has a private key. In this paper, we propose an RSA-based PAKR (for short, RSA-PAKR) protocol using multiple n servers, and then show that the RSA-PAKR protocol provides security of passwords and static keys against an attacker who takes full control of n - 1 servers. Several aspects of the RSA-PAKR protocol and its extension (including efficiency improvements, implementation issue and extension for distinguishing on-line dictionary attacks) are followed.