AI-Driven Vulnerability Management and Automated Threat Mitigation
Venkata Bhardwaj Komaragiri, Andrew Edward
Abstract
Corporate networks face an increasing number, diversity, and sophistication of threats. Classical perimeter security mechanisms are less effective or even counter-productive. Continuous vulnerability management becomes a crucial task. AI-based machine learning and data mining techniques have been shown to improve vulnerability prediction and prioritization. To close the feedback loop, they can also be applied to evaluate and mitigate identified vulnerabilities by predicting associated risk scores and suggesting remediation measures. Overall, AI-driven vulnerability management can automate most steps of the Risk Management Framework prescribed by the United States National Institute of Standards. While AI-based prediction models and tools are evaluated on datasets such as the Common Vulnerability Scoring System (CVSS), performance evaluation is limited by the available data. More complex risk prediction models require richer datasets and, thus, involve machine learning models that are more demanding on enterprise data privacy policies and, consequently, are difficult to assess. Still, AI-driven threat management can ensure that the "security rodent race" is in favor of the enterprise.