Side-channel attacks are commonly used to extract cryptographic keys from embedded systems. We study side-channel countermeasures originally designed to protect against differential power analysis (DPA) and determine how effective they are against profiling attacks using neural networks. We compiled two otherwise identical binaries that perform AES-256 encryption, one with DPA protections and one without. In our experiments, the binary with the countermeasures required 8 to 18 times as much attack data as the unprotected binary to achieve the same efficacy of attack, given the same amount of training data for both binaries. We conclude that our DPA protections may make profiling attacks time prohibitive in some cases but they are insufficient to completely defend against profiling attacks.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Robinson, Arthur

Subject

• Computing
• Security
• Innovation

Publisher

• Worcester Polytechnic Institute

Identifier

• 41201
• E-project-110821-150024

Advisor

• Walls, Robert Joseph

Year

• 2021

Date created

• 2021-11-08

Resource type

• Major Qualifying Project

Major

• Computer Science

Rights statement

• In Copyright