A Security-Aware Side-Channel Detection Through Convolutional Transformer Networks and Hybrid LSTM-Spectral Analysis

Networks and Hybrid LSTM-Spectral Analysis

Abstract

This paper introduces a novel method for detecting side-channel attacks in embedded system security by combining Long Short-Term Memory (LSTM), spectral analysis, and convolutional transformer networks. The proposed hybrid model achieves 97% detection accuracy, 95% precision, and 96% recall. It integrates Convolutional Neural Networks (CNN) for spatial information extraction, Transformers for sequence data modeling, LSTM for analyzing temporal details, and spectral analysis for frequency domain insights. The framework demonstrates robust real-time performance in identifying subtle side-channel leakages, providing an effective solution for addressing security challenges in critical systems.

Objectives: The main objectives of this research are to develop an advanced side-channel attack detection system for embedded security by utilizing a hybrid model that combines spatial, temporal, and frequency domain analyses. The model also aims to provide high detection accuracy and real-time performance to offer a comprehensive solution for cybersecurity challenges.

Methods: The proposed method employs a hybrid approach combining CNNs for spatial data extraction, Transformers for sequence modeling, LSTM units for temporal investigation, and spectral analysis for insights from the frequency domain. This combination ensures accurate detection of side-channel attacks in embedded systems, with a focus on real-time applicability.

Results: The hybrid model delivers a detection accuracy of 97%, precision of 95%, and recall of 96%. The integration of CNN, Transformer, LSTM, and spectral analysis offers robust performance in detecting subtle side-channel attacks, outperforming conventional methods in terms of detection rates and real-time capability.

Conclusion: This hybrid LSTM-Spectral Analysis and convolutional transformer network model provides an efficient and comprehensive solution to detecting side-channel attacks in embedded systems. Its high accuracy, precision, and recall, combined with real-time detection capability, make it a superior approach for enhancing security in critical systems facing evolving cyber threats.