Acoda: Adversarial Code Obfuscation for Defending against LLM-based Analysis

2026-06-10

AI Summary

This work addresses the security and privacy risks posed by large language models (LLMs) in code understanding tasks, where models may inadvertently expose source code logic and intellectual property. The authors propose Acoda, a novel framework that leverages intrinsic LLM mechanisms—such as safety alignment and token-level processing—to generate semantically preserving adversarial code obfuscation strategies via genetic algorithms. These strategies are iteratively optimized to disrupt LLMs’ accurate code analysis while strictly maintaining original program semantics. Acoda achieves high cross-model transferability and minimal runtime overhead. Furthermore, the study introduces a multi-dimensional evaluation protocol based on auxiliary LLMs to quantitatively assess code comprehension capabilities. Experiments demonstrate attack success rates of up to 70% across seven prominent LLMs, effectively inducing denial-of-service or misjudgment without altering functional behavior.

Abstract

With the widespread adoption of Large Language Models (LLMs) in software engineering (SE) tasks such as code understanding, debugging, and vulnerability detection, their powerful semantic reasoning ability has also introduced new security and privacy risks. LLMs can analyze, reconstruct, or even reverse-engineer source code logic, potentially leading to the leakage of intellectual property. To address this issue, we propose Acoda, a genetic algorithm-based adversarial code obfuscation framework that defends against LLM-based code analysis. Acoda leverages two key mechanisms of LLMs, namely safety alignment and token-based information processing, to design 8 semantics-preserving obfuscation methods. It iteratively optimizes obfuscation strategies through a genetic algorithm to generate adversarial samples that maximize defensive effectiveness. In addition, we propose a quantitative evaluation framework based on LLM responses, which combines an auxiliary LLM and four evaluation metrics to assess how target LLMs analyze obfuscated code comprehensively. Experimental results show that Acoda can effectively induce LLMs to refuse or misinterpret code analysis. On 7 state-of-the-art LLMs, including GPT-4o, DeepSeek, Qwen, Llama, and Gemma, Acoda achieves an attack success rate (ASR) of up to 70%, with strong cross-model transferability and minimal runtime overhead, while ensuring that the semantics of the original code remain unchanged. Overall, this study provides a new perspective for code protection and LLM security defense in the era of LLMs.

Problem

• Adversarial Code Obfuscation
• LLM-based Analysis
• Code Privacy
• Intellectual Property Protection
• Semantic Preservation

Innovation

• adversarial code obfuscation
• genetic algorithm
• LLM security
• semantic-preserving transformation
• quantitative evaluation framework