Date of Award

6-6-2025

Document Type

Thesis

Departments

Computer Science and Engineering; Electrical and Computer Engineering

First Advisor

Hoeseok Yang

Second Advisor

Younghyun Cho

Abstract

Differential Fault Analysis (DFA) is a potent hardware attack that threatens cryptographic security by injecting faults into a cipher implementation to reveal secret keys. This project aims to mitigate DFA attacks on the Advanced Encryption Standard (AES) by implementing targeted countermeasures in an embedded AES-128 encryption core. Two key techniques are explored: Randomization and Triple Modular Redundancy (TMR). The randomization approach introduces unpredictability into the encryption process, which involves inserting dummy rounds, artificial noise, and random delays, to disrupt an attacker’s timing and analysis, while TMR provides redundancy by replicating critical rounds of computation and using majority voting to correct any single-fault errors. The effectiveness of these countermeasures was evaluated using a real fault-injection attack scenario: a clock glitch was used to induce faults in the AES encryption process, and the outputs were analyzed for key leakage. Results show that with the countermeasures in place, the DFA attack was unable to recover the AES secret key, whereas an unprotected AES implementation succumbed to key extraction. These findings highlight the importance of integrating hardware-level defenses to secure cryptographic devices against fault injection attacks, achieving improved security with minimal performance trade-offs.

Download

Share

COinS