This research introduces novel approaches to address privacy vulnerabilities in code-generating AI models by removing memorized sensitive information.

• Tackles the memorization problem in LLMs for code that can leak sensitive data embedded during training
• Proposes machine unlearning techniques specifically designed for code generation models
• Preserves model utility while enhancing privacy protection mechanisms
• Establishes evaluation frameworks to measure effectiveness of unlearning methods

This work is critical for security professionals as it addresses compliance requirements and builds user trust by preventing unwanted data exposure in AI-powered development tools.

Mitigating Sensitive Information Leakage in LLMs4Code through Machine Unlearning