LATIDIA · Ciberseguridad

La decodificación restringida por la gramática puede hacer Jailbreak a los LLM para generar código malicioso

arXiv: 2606.11817v3Tipo de Anuncio: reemplazar Resumen: Los Modelos de Lenguaje Grande (LLM) se utilizan cada vez más para la generación de código, lo que plantea la preocupación de que puedan ser mal utilizados para producir código malicioso. Mientras tanto, Grammar-Constrai

La noticia

arXiv:2606.11817v3 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly used for code generation, raising concerns that they may be misused to produce malicious code. Meanwhile, Grammar-Constrained Decoding (GCD) has been widely adopted to improve the reliability of LLM-generated code by enforcing syntactic validity. In this paper, we reveal a counterintuitive risk: this reliability-oriented technique can itself become an attack surface. We uncover a new jailbreak attack, termed CodeSpear, that exploits GCD to induce LLMs into generating malicious code. Our experiments show that simply applying a benign code grammar constraint can effectively jailbreak LLMs. To address this vulnerability, we propose CodeShield, a safety alignment approach that robustly preserves safe