LLMInjector: Automating Prompt Injection Vulnerability Detection for Large Language Models in Burp Suite

LLMInjector is a Burp Suite extension tool designed specifically for security testers, aiming to automate the discovery and analysis of prompt injection security risks in LLM integrations. It can be used without programming background, helping identify security weaknesses in AI-driven systems by simulating various prompt manipulation attacks, lowering the technical barrier for AI security testing, and assisting enterprises and researchers in systematically verifying the security of AI systems.

As large language models (LLMs) are increasingly integrated into various applications, prompt injection attacks have become one of the severe security challenges facing AI systems. Attackers can bypass model security restrictions, steal sensitive information, or manipulate model behavior through carefully crafted inputs. For security testers, systematically detecting such vulnerabilities has become an urgent task.

LLMInjector as a Burp Suite extension has the following core features:

1. Seamless Burp Suite Integration: Intercept requests in the Proxy tab and directly run injection tests on AI application endpoints;
2. Comprehensive Attack Simulation: Supports automated testing of multiple types such as direct/indirect prompt injection and jailbreak attacks;
3. Clear Security Reports: Generates reports containing potential flaws to help quickly locate issues and assess risks;
4. Zero-Code Operation: Configure and execute tests via a graphical interface, lowering the technical barrier.

Operating Environment

• Operating System: Windows 10+ (64-bit)
• Java Environment: JRE 11+
• Burp Suite: Professional/Community Edition (2021.3+)
• Memory: At least 2GB of available RAM
• Network: Internet connection required for installation and updates

Installation Steps

1. Download the LLMInjector .jar file
2. Open Burp Suite and go to the Extender tab
3. Select the Extensions sub-tab and click Add
4. Choose the Extension type as Java
5. Load the downloaded .jar file

LLMInjector's main use cases include:

1. Enterprise Security Testing: Verify the security of AI components before product release to reduce risks;
2. Security Research: Explore the principles of prompt injection attacks and new attack vectors, contributing to defense ideas;
3. Compliance Verification: The generated test reports can serve as supporting materials for AI system security compliance.

Notes for using LLMInjector:

• Only use for authorized security testing;
• Evaluate risks of test results in combination with business scenarios;
• The field of prompt injection is evolving rapidly, so the tool needs regular updates to address new types of attacks.

LLMInjector fills the gap in the field of AI security testing tools, encapsulating complex prompt injection detection technology into an easy-to-use Burp extension, allowing more security personnel to participate in AI system protection. As LLM applications become more widespread, such specialized security testing tools will become increasingly important.