Broken Access Control Detection Focused on Privilege Escalation Prevention Using a Llama 3 LLM-Based Assistant

Broken Access Control (BAC) found itself atop the Open Worldwide Application Security Project (OWASP) top 10 risks of 2021. Cementing the importance and danger of this cyber threat. Access control enforces policies that ensure users are only able to access what they are permitted to. Failures within access control lead us to BAC, which introduces a variety of security hazards such as data breaches and leaks. In a society where data is becoming invaluable, having a cyber threat of this caliber is potentially detrimental to businesses and consumers alike. Utilizing the ever-evolving field of artificial intelligence (AI) to combat BAC can revolutionize the way security measures are created. Implementing an AI assistant to monitor and detect BAC during the developmental process to resolve potential BAC before an attack occurs allows for a proactive rather than a reactive approach. This study will specifically focus on privilege escalation, a key component and leading cause in BAC, and on the Llama 3 Large Language Model (LLM) from Meta. Llama was chosen due to its role as a pioneer within the open-source LLM community, rivaling the products of OpenAI's ChatGPT and Anthropic's Claude.