A Framework for Robustness Evaluation in AI-Based Malware Detectors

This paper presents a framework for the robustness evaluation of AI-based malware detectors. The framework is composed of three elements: Training phase, Evaluation phase, and Attack experimental. Researchers upload their own dataset, feature extraction scrips, or trained model in the framework for evaluation. The evaluation metrics measure the accuracy, F-1, false positive, and false negative rate of the detectors, and evaluation results are used to confirm effectiveness and e ciency. Our proposed framework aids in resolving the issue that researchers don’t need to possess an entire experimental setting, especially considering that some research sources are di cult to collect and integrate.