Metrics Gathering for AI-based Trust Assessment

Technological advances have increased the complexity and volume of system operations. As a result, today systems face broader attack surfaces due to the larger amount of code executed. Ensuring the reliability of applications and operations in such environments requires well-defined security criteria, but establishing them is a non-trivial task. Trusted Computing offers mechanisms to establish trust in a non-repudiable manner, yet it faces inefficiencies and scalability challenges when applied to complex scenarios. In fact, defining a flexible verification method for a complex system is a challenging task. This is especially true when the process focuses solely on verifying individual actions rather than on the overall system’s behaviour. AI-based techniques offer a viable countermeasure, since modelling a system’s behaviour provides an adaptable approach to determining reliability. An AI model can also analyse a larger amount of data. However, collecting, securely storing, and provisioning relevant information is still a required feature. This work proposes two approaches for the secure collection of system events. The objective is to facilitate the provision of an AI model while ensuring the integrity and authenticity of the data.