Quote of the day

A person who never made a mistake never tried anything new.

- Albert Einstein

Contact

Defending Agentic LLM Systems against Prompt Injection and Tool Abuse

Open Cybersecurity

Defending Agentic LLM Systems against Prompt Injection and Tool Abuse

Defending Agentic LLM Systems against Prompt Injection and Tool Abuse

Description

Build an evaluation framework and programmable privilege-control mechanisms to reduce the impact of prompt injection and tool abuse in LLM-based agents.

Details

Context and Problem Statement

LLM-based agents consume untrusted content such as web pages, emails, and documents while accessing tools capable of acting on external systems. Instructions embedded in data can therefore redirect the agent toward unintended actions, while text-level filtering remains fragile against indirect attacks.

Research Question

Does programmable privilege control at the tool layer reduce the attack surface more effectively than prompt filtering at the text layer, and what functional cost does such protection impose?

Proposed Approach

Design a mediation layer between the agent and its tools that enforces least privilege by task, human confirmation for irreversible actions, and isolation of untrusted data. A structured attack benchmark measures compromise rate, false blocking, and utility degradation.

Expected Contribution

A comparative evaluation of defense families using a joint security-utility metric on a reproducible benchmark.

Expected Prototype

A reusable tool-mediation library with automated red-team scenarios and a coverage report.

Datasets

Published prompt-injection scenarios and an agent-task corpus built for isolated experimentation.

Challenges

Defining data-instruction boundaries, combinatorial attack diversity, unstable benchmarks, and the trade-off between security and usability.

Research Question

Which combination of privilege control and data isolation most effectively reduces prompt-injection risk without undermining agent utility?

Innovation

The project moves defense from the fragile textual layer to architectural privilege control and explicitly measures the security-utility trade-off.

Expected Deliverable

An open-source mediation layer, an injection-attack benchmark, and a comparative study of defense mechanisms.

Technologies

LLM Agents and Tool Use Access Control and Declarative Policies Sandboxing and Isolation Automated Red Teaming Trustworthy AI

Required Skills

  • Application Security and Threat Modeling
  • Python Software Development and Agent Architecture
  • Access-Control and Policy-Verification Fundamentals
  • Experimental Methodology

Datasets

  • Prompt-injection scenarios from academic publications
  • Agent task corpus developed for controlled experimentation

Morocco & Africa Relevance

Moroccan organizations are rapidly adopting tool-enabled AI assistants without mature security architectures. Reusable guardrails address an immediate regional need.