Enhancing Secret Detection in Cybersecurity with Small Language Models

Explore the innovative application of small language models for secret detection in cybersecurity through this 31-minute conference talk from BSidesSF 2025. Learn how Danny Lazarev and Erez Harush tackle the challenges of using large language models in cybersecurity by demonstrating the fine-tuning process of an LLM specifically designed for detecting secrets in code while maintaining efficiency for laptop deployment. Discover whether low-inference-time LLMs can unlock previously overlooked detection methods and understand the practical implications of implementing lightweight AI solutions for cybersecurity applications. Gain insights into the balance between model performance and computational efficiency in real-world security scenarios.