Darktrace funded studentships -01

Artificial Intelligence (AI) systems, particularly Large Language Models (LLMs), have demonstrated significant capabilities across diverse application domains. AI agents, autonomous software programs driven by LLMs that can perceive their environment, make decisions, and take actions to achieve goals, have started to be widely used to solve problems in traffic, medicine, research, and software development. Communication between agents and traditional systems has become a significant challenge. MCP (Model Context Protocol) is an open-source standard for connecting AI agents to external systems, including other AI agents, introduced in late 2024 and quickly became a common standard. With MCP, external systems serve as MCP servers to provide services to these AI agents. However, it has been demonstrated that malicious MCP servers can mislead AI agents and even instruct them to perform malicious activities on users' computers, posing a critical issue for AI-agent-based systems. As MCP has only recently emerged, effective methods for identifying malicious MCP servers are not yet available. This project aims to identify potentially malicious MCP servers through analysis of their open-sourced develop defence mechanisms to protect AI agents from compromise. We will develop a multi-agent system (MAS) for program analysis, in which multiple autonomous agents collaborate to analyse software, requiring the development of coordination methods. All analysis will be conducted using open-source code and publicly available benchmark data, ensuring reproducibility and enabling publication of findings. 94% of Loughborough’s research impact is rated world-leading or internationally excellent. REF 2021

The standard University IELTS English language requirement is 6.5 overall with 6.0 in each individual element (reading, writing, listening and speaking).

October 2026

21 July 2026

• Loughborough, United Kingdom