Analysis of improved side-channel attack on AES-128 with masking

• Main Author: Tee, Yee Yang
• Other Authors: Gwee Bah Hwee
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering::Integrated circuits
• Online Access: https://hdl.handle.net/10356/141888
• Institution: Nanyang Technological University

Encryption is essential for data security in the information age today. The Advanced Encryption Standard (AES) is regarded to be the global ‘gold’ standard for secure encryption. Today, many embedded processors like smart cards are used to store to secret keys, and these are vulnerable to cyber-physical attacks like Side Channel Attack (SCA). Adversaries can exploit the power dissipation, electromagnetic emanation or timing information during the encryption process and obtain the secret key. Due to the possibilities of SCA, many processors today have protective measures like masking and hiding implemented.Masking is used to decorrelate power consumption and reduce information leakage. An improvement to counter masking techniques is a second order attack, which performs the attack on 2 separate points in the power trace. This research focuses on the effectiveness of second order Correlation Power Analysis (CPA), on a masked implementation of AES-128. CPA was first conducted on an unmasked AES-128 design to confirm its effectiveness. This was successful with 1500 power traces. It was also tested against a masked implementation, which proved that masking is highly effective against standard CPA attacks, as no key bytes were obtained with more than 10,000 power traces. Subsequently, second order attacks were conducted against the same masked implementation, which also proved to be secure against such attacks. The masked AES was resilient against second order attacks using more than 10,000 traces, almost 800% of the amount required for an unmasked design. Signal-to-Noise Ratio was used to determine the leakage from changing mask values, and it was concluded that power consumption caused by mask generation was too insignificant to be exploitable.