How machine learning enables automated side-channel detection / Jan Peter Drees, M.Sc. [Wuppertal], April 19, 2023
Inhalt
- Introduction
- Fundamentals
- New Techniques for Side-Channel Detection with Machine Learning
- Introduction
- The Information Leakage Problem
- Our Approaches for Information Leakage Detection
- Empirical Evaluation
- Conclusion and Open Problems
- Application to Bleichenbacher's Attack
- Introduction
- Related Work
- Preliminaries
- Implementation of Automated Side-Channel Detection
- Manipulated TLS Client
- Feature Extraction
- Classification Model Learning
- Error Correction and Report Generation
- Analysis
- Test Setup
- Basic Approach Validation
- Detecting Klíma-Pokorný-Rosa Side Channels
- Detecting ROBOT Side Channels
- Testing open-source Implementations
- Commercial Integration
- Conclusions and Open Problems
- Automating Hardware Attacks
- Introduction
- Related Work
- Background
- Supervised Learning for Profiled Side-Channel Attacks
- Convolutional Neural Networks
- Leakage Model
- Neural Architecture Search
- Our Approach
- Setup of Our Parameter Study
- Parameter Study Results
- Overall Reliability
- Optimal Neural Architecture Search Parameters
- Comparison with Fixed Architectures
- Conclusion and Open Problems
- Conclusion and Outlook
- Bibliography
- Appendix