DATDP (Defense Against The Dark Prompts) offers a breakthrough solution to the security vulnerability of Best-of-N jailbreaking attacks that plague modern language models.

• Achieves 100% effectiveness in blocking jailbreaks from the original BoN paper
• Blocks 99.8% of jailbreaks in independent replication tests
• Detects random augmentations like unusual capitalization and punctuation used to manipulate LLMs
• Provides robust protection without compromising normal model functionality

This research significantly advances LLM security by addressing a critical vulnerability that affects all major language models, making AI deployments substantially safer in real-world applications.

Defense Against the Dark Prompts: Mitigating Best-of-N Jailbreaking with Prompt Evaluation