The evolution of Large Language Models (LLMs) and their increased reasoning capabilities have transformed the cybersecurity landscape, as cybercriminals actively use them to detect vulnerabilities at high speed. This adoption of artificial intelligence by attackers highlights the fact that traditional auditing tools, based on static rules and signatures, are insufficient to deal with dynamic and adaptive threats.
Faced with this challenge, inLab FIB is working on the development of an advanced pentesting system based on a multi-agent architecture that overcomes the limitations of conventional automation. The goal is to have a tool capable of understanding the logic of the audited systems, speeding up the discovery of vulnerabilities and drastically reducing the total time needed to complete an audit, while following a clear and established methodology. In this way, the system assumes an initial analysis, allowing the auditor to focus on discovering the most complex vulnerabilities.
One of the elements taken into account during the development of the project is the privacy and sovereignty of the data. For this reason, the architecture of the system has been designed to be completely flexible: it will have the capacity to operate with both commercial model providers and local open source models. The latter modality guarantees that no confidential data reaches third-party servers, thus complying with data protection standards.
–
