MCP tools are implicated in several new attack techniques. Here's a look at how they can be manipulated for good, such as logging tool usage and filtering unauthorized commands.
Over the last few months, there has been a lot of activity in the Model Context Protocol (MCP) space, both in terms of adoption as well as security. Developed by Anthropic, MCP has been rapidly gaining traction across the AI ecosystem. MCP allows Large Language Models (LLMs) to interface with tools and for those interfaces to be rapidly created. MCP tools allow for the rapid development of “agentic” systems, or AI systems that autonomously perform tasks.
Beyond adoption, new attack techniques have been shown to allow prompt injection via MCP tool descriptions and responses, MCP tool poisoning, rug pulls and more.
Prompt Injection is a weakness in LLMs that can be used to elicit unintended behavior, circumvent safeguards and produce potentially malicious responses. Prompt injection occurs when an attacker instructs the LLM to disregard other rules and do the attacker’s bidding. In this blog, I show how to use techniques similar to prompt injection to change the LLM’s interaction with MCP tools. Anyone conducting MCP research may find these techniques useful.
This vulnerability can allow attackers to steal anything a user puts in a private Slack channel by manipulating the language model used for content generation. This was responsibly disclosed to Slack (more details in Responsible Disclosure section at the end).
Like many companies, Dropbox has been experimenting with large language models (LLMs) as a potential backend for product and research initiatives. As interest in leveraging LLMs has increased in recent months, the Dropbox Security team has been advising on measures to harden internal Dropbox infrastructure for secure usage in accordance with our AI principles. In particular, we’ve been working to mitigate abuse of potential LLM-powered products and features via user-controlled input.
Large Language Models (LLM) have made amazing progress in recent years. Most recently, they have demonstrated to answer natural language questions at a surprising performance level. In addition, by clever prompting, these models can change their behavior. In this way, these models blur the line between data and instruction. From "traditional" cybersecurity, we know that this is a problem. The importance of security boundaries between trusted and untrusted inputs for LLMs was underestimated. We show that Prompt Injection is a serious security threat that needs to be addressed as models are deployed to new use-cases and interface with more systems.
[PDF DOC] https://arxiv.org/pdf/2302.12173.pdf
