PR29 INCYBER Can AI help us understand vulnerabilities Exploring AI driven approaches to cybersecu
In this masterclass, we’ll explore how artificial intelligence (AI) is transforming the understanding of software vulnerabilities to strengthen cybersecurity. We will begin by examining how vulnerabilities are typically disclosed, with descriptions detailing their exploitation by malicious actors. As vulnerability databases continue to expand, efficiently analyzing and prioritizing them has become a growing challenge. This is where AI comes in—by leveraging language models (LMs), we can automate and enhance vulnerability analysis at scale. LMs are powerful AI tools that learn from text, making them ideal for processing vulnerability descriptions and automating security tasks. This session will provide a brief introduction to how LMs work, how they can be fine-tuned for new tasks, and explore applications in cybersecurity. One key application is automating risk assessment during the disclosure process: LMs can predict the tactics that attacker can leverage from vulnerabilities based on their descriptions, enabling security teams to prioritize threats more effectively. Another innovative application will be explored: using AI agents as cyberattackers to discover the most critical attack paths within a network. An AI agent can analyze a network with known vulnerabilities and simulate an attacker’s decision-making, strategically selecting and sequencing vulnerabilities to exploit based on a specific threat model. By analyzing the identified attack paths, organizations can pinpoint critical weaknesses and implement proactive defenses before real threats emerge.
Allez à la source