HW/SW Co-Design for Security Systems and the Investigation of Deep Learning-based Side-channel Analysis

Abstract

Electronic devices have permeated into all aspects of our lives, from basic smart cards to sophisticated hybrid automobile systems. These devices comprise a range of products like sensors, wearable gadgets, mobile phones, personal computers, and others, playing vital roles in many applications and enabling the Internet of Things (IoT). However, with this interconnectedness comes the associated security risks since attackers can exploit vulnerabilities in the system.

Securing electronic devices requires the use of cryptographic algorithms and trusted execution environments (TEEs). Cryptographic algorithms ensure data confidentiality and integrity through encryption/decryption, hashing, and digital signatures. TEEs provide secure enclaves within the system for critical operations that prevent unauthorized modifications and access by imposing stringent access restrictions. These two measures have become robust mechanisms for enhancing the security of critical operations and data access control.

Despite the above security measures, electronic systems are susceptible to various attacks, including side-channel analysis (SCA), in which attackers exploit information leakage from physical devices while executing instructions or cryptographic algorithms. Power consumption and electromagnetic radiation (EM) are common indicators of this leakage. Countermeasures such as masking and hiding techniques are commonly employed to enhance resistance against SCA. However, the advent of deep learning in SCA has brought forth new challenges, rendering previously efficient countermeasures ineffective. Moreover, deep learning-based SCA has the potential to eliminate preprocessing and alignment requirements inherent in earlier methods.

Therefore, this thesis focuses on two main objectives. The first objective is the implementation of cryptographic algorithms and the incorporation of TEEs for secure-sensitive applications. HW/SW co-design approach will be utilized to attain optimal performance while preserving flexibility. The second objective of this thesis is the investigation of deep learning-based SCA to explore its effectiveness in detecting side-channel vulnerabilities.