Inline malicious url detection with hierarchical structure patterns

Abstract

A hierarchical structure constructor constructs a hierarchical structure that comprises nodes associated with feature sets patterns of URLs. Nodes at each depth are labelled as malicious, benign, or mixed for corresponding to URLs that are malicious, benign, or malicious and benign that match the corresponding patterns. Malicious feature set patterns are extracted from malicious nodes in the hierarchical structure. A URL analyzer operates inline by logging traffic sessions, extracting URLs from the logs, and matching the extracted URLs with the malicious feature sets patterns extracted from the hierarchical structure. The hierarchical structure is periodically updated with known malicious/benign URLs to improve quality of malicious URL detection.