Examining Adversarial Learning against Graph-based IoT Malware Detection Systems
It addresses security vulnerabilities in IoT malware detection systems, which is critical for protecting IoT devices, but the work is incremental as it applies known adversarial techniques to a specific domain.
This study investigated the robustness of graph-based deep learning models for IoT malware classification against adversarial learning, finding that off-the-shelf attacks achieved a 100% misclassification rate and a graph embedding approach misclassified all malware samples as benign.
The main goal of this study is to investigate the robustness of graph-based Deep Learning (DL) models used for Internet of Things (IoT) malware classification against Adversarial Learning (AL). We designed two approaches to craft adversarial IoT software, including Off-the-Shelf Adversarial Attack (OSAA) methods, using six different AL attack approaches, and Graph Embedding and Augmentation (GEA). The GEA approach aims to preserve the functionality and practicality of the generated adversarial sample through a careful embedding of a benign sample to a malicious one. Our evaluations demonstrate that OSAAs are able to achieve a misclassification rate (MR) of 100%. Moreover, we observed that the GEA approach is able to misclassify all IoT malware samples as benign.