Less Is More: Robust and Novel Features for Malicious Domain Detection

Malicious domains are increasingly common and pose a severe cybersecurity threat. Specifically, many types of current cyber attacks use URLs for attack communications (e.g., C&C, phishing, spear-phishing). Despite the continuous progress in detecting attacks, there still critical weak spots structure defense mechanisms. Since machine learning has become one most prominent malware detection methods, robust feature selection mechanism is proposed that results malicious domain models resistant to evasion attacks. This exhibits high performance based on empirical data. paper makes two main contributions: First, it provides an analysis widely used features literature. Note even though set dimensional space cut by half, classifier improved (an increase model’s F1-score from 92.92% 95.81%). Second, introduces novel with regard adversary’s manipulation. Based extensive evaluation different sets commonly classification models, this shows perturbations concurrently helpful classifying non-manipulated