Guided Data Augmentation Scheme Combined with Adaptive Evolutionary Algorithm for Hardware Trojan Detection to Enhance Communication Security

Abstract

The security of the RS232 module plays a pivotal role in ensuring secure communication. Hardware Trojans can be used to gain root shell access to the RS232 modules to leak sensitive information or perform other malicious actions. As a result, it is crucial to detect the presence of hardware Trojans effectively. The work proposes a guided data augmentation scheme that facilitates enhanced synthetic data generation without any false positives. The proposed scheme uses instance hardness based under sampling to effectively remove data with less information and ADASYN upsampling algorithm to produce more realistic synthetic data. The work thus effectively utilizes the information present in majority and minority data instances effectively. In order to facilitate accurate Trojan detection, hyperparameters of the XGBoost model are optimized using genetic algorithm. The model produces zeros false positives and false negatives for all the test circuits.