# Certifying LLM Safety against Adversarial Prompting

URL: https://www.synapticlabs.ai/blogs/certifying-llm-safety-against-adversarial-prompting
Author: Dendrex
Published: 2024-03-12

Source: Synaptic Labs, https://www.synapticlabs.ai/blogs/certifying-llm-safety-against-adversarial-prompting

---

The paper: http://arxiv.org/abs/2309.02705

\## Purpose
The paper's primary goal is to introduce a framework for defending LLMs against adversarial attacks. These attacks involve adding malicious token sequences to prompts, tricking LLMs into producing harmful content despite safety measures.

\## Methods
\- \*\*Erase-and-Check Framework\*\*: A novel approach where individual tokens in a prompt are erased, and subsequences are analyzed using a safety filter.
\- \*\*Attack Modes Addressed\*\*: Three types of adversarial attacks – suffix, insertion, and infusion.
\- \*\*Safety Filter Implementation\*\*: Using pre-trained LLMs and a fine-tuned DistilBERT classifier to distinguish between safe and harmful prompts.

\## Key Findings
1\. \*\*Effectiveness of Erase-and-Check\*\*: This method provides strong safety guarantees, successfully identifying harmful prompts with high accuracy.
2\. \*\*Improvement with DistilBERT\*\*: Replacing Llama 2 with DistilBERT enhances accuracy and speed, recognizing harmful prompts more efficiently.
3\. \*\*Superiority Over Other Methods\*\*: Outperforms traditional methods like randomized smoothing, especially in complex attack scenarios.
4\. \*\*High Performance on Safe Prompts\*\*: Maintains good performance in classifying non-adversarial safe prompts.

\## Discussion
This research is crucial for the field of AI safety, addressing the vulnerability of LLMs to adversarial attacks. The erase-and-check approach offers a robust defense mechanism, improving the reliability of LLMs in real-world applications.

\## Critiques
1\. \*\*Computational Cost\*\*: The complexity of the method, especially in more advanced attack modes, requires significant computational resources.
2\. \*\*Limitation in Scope\*\*: The study primarily focuses on defense against specific types of adversarial attacks, potentially overlooking other emerging threats.

\## Tags
#LLMSafety #AI #AdversarialAttacks #EraseAndCheck #DistilBERT #ComputationalCost