Mitigating Generative Artificial Intelligence (AI) Cybersecurity Risks Leveraging Retrieval Augmented Generation (RAG) in Telemetry Post-Processing Analysis

Abstract

Using AI tools to assist in telemetry post processing analysis can offload tedious analysis tasks as well as provide opportunities for gaining better and new insights into interpreting data. However, deploying generative AI applications into sensitive telemetry post processing environments can introduce significant cybersecurity risks; specifically, with respect to generative AI models potentially integrating sensitive program data into the underlying model being used. If configured properly, Retrieval Augmented Generation (RAG) provides a capability for generative AI solutions to leverage important information that will not be integrated into the fundamental AI model being utilized. After reviewing the current generative AI threat landscape, examples will be given regarding how RAG may be employed to better mitigate generative AI cybersecurity risks in telemetry post processing environments.