EM Side-Channel Based System Monitoring and Profiling with Neural Networks
Designed and trained neural networks to profile programs running on IoT devices by exploiting electromagnetic (EM) signals emitted during execution.
Detected malware by identifying anomalies in program behavior using EM side-channel data, enhancing system security.
Achieved non-intrusive classification of circuit components without direct physical access, advancing remote diagnostic capabilities.
Conducted experiments and collected data using a variety of signal-capturing devices, including Software-Defined Radios (SDRs) and spectrum analyzers, to ensure high-quality signal acquisition and analysis.
Channel Modelling and Instruction Tracking
Developed models to understand and analyze input signals leading to EM emissions during software activity.
Constructed instruction-specific EM signal models for different devices, improving accuracy in signal interpretation.
Demonstrated instruction-level tracking of scripts executed by computer systems using advanced signal processing techniques.
Quantifying Information Leakage Caused by Software Activities
Analyzed and quantified potential information leakage from computing devices caused by software activity.
Demonstrated the correlation between instruction sequence, implementation, and the extent of information leakage.
Established mathematical bounds on information leakage, providing a theoretical framework for side-channel mitigation.
Bridged conventional communication systems with covert/side channels to better understand their behavior and impact on security.
OpenSSL Cryptography Vulnerability Analysis
Applied advanced signal-processing techniques to extract sensitive information from EM emissions during OpenSSL signing operations.
Identified and disclosed cryptographic vulnerabilities in OpenSSL to strengthen its resistance to EM side-channel attacks.
Proposed countermeasures to mitigate EM-based threats, contributing to the enhancement of cryptographic security.
Compressed Equalization
Developed compressed equalization techniques to exploit hidden sparsity in combined channel responses for improved signal reconstruction.
Derived error bounds between actual and estimated combined channels, enhancing the robustness of the equalization process.
Conducted experimental verification with MATLAB simulations for both fractionally and time-spaced equalization scenarios, demonstrating practical feasibility.
Bounded Component Analysis
Developed novel Bounded Component Analysis (BCA) methods and algorithms to separate dependent and independent sources from their instantaneous mixtures.
Validated the proposed algorithms through experimental MATLAB simulations, achieving effective source separation in real-world scenarios.