Skip to content

andthemomentoflife/sentry

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 

Repository files navigation

SENTRY: Defense Mechanisms Against LLM Jailbreak Attacks

This repository contains a course project for Introduction to Information Security. The project surveys existing defense mechanisms designed to mitigate jailbreak attacks in Large Language Models (LLMs) and organizes them into a structured taxonomy.

The accompanying paper analyzes how current research addresses jailbreak vulnerabilities and categorizes defense strategies based on when they intervene in the LLM pipeline.


Paper

📄 Full Paper

View the Paper (PDF)


Overview

Large Language Models such as GPT-4, Bard, and LLaMA are powerful text generation systems, but they remain vulnerable to jailbreak attacks. Jailbreak attacks attempt to bypass safety mechanisms through adversarial prompts, leading models to generate harmful or restricted content.

This project reviews existing research on defense mechanisms against jailbreak attacks and categorizes them into two main groups:

1. Pre-processing Defenses

These approaches modify or strengthen the model before user interaction.

Examples include:

  • Fine-tuning models with safety datasets
  • Generating adversarial training examples
  • Selecting specialized model architectures based on prompt characteristics

2. Mid / Post-processing Defenses

These approaches intervene during or after the model processes a user query.

Examples include:

  • Prompt template transformations
  • Self-reminder or safety prompts
  • Token-level probability manipulation during decoding

Key Contributions

  • Survey of recent research on LLM jailbreak defense mechanisms
  • A taxonomy of defense strategies based on intervention stage
  • Comparative discussion of advantages and limitations of each category
  • Identification of future research directions for improving LLM safety

Topics Covered

  • LLM security and safety
  • Jailbreak attacks
  • Defense mechanisms
  • Prompt-based safety techniques
  • Token-level decoding defenses

Course Information

Course: Introduction to Information Security Institution: Korea University Semester: 2024-2


Authors

  • Miryeong Kang
  • Minjae Kang

About

Secure Exploration of Neutralizing Threats and Responses Yielded

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors