Submit #795257: Bytedance verl <=0.7.0 Arbitrary Code Execution

verl is an open-sou large model reinfoment learning training framework by ByteDance (20k+ Stars), supporting algorithms such as PPO and GRPO. In its math answer scoring module prime_math/grader.py, the math_equal() function is used to compare whether the model-generated answer is equivalent to the ground truth answer.

When the ground truth answer is a matrix type (containing \begin{pmatrix}) and the model's output answer starts with [ and ends with ], the code directly calls Python's built-in eval() function on the model output without any input sanitization or sandbox isolation.

An attacker can use Indirect Prompt Injection (injecting malicious instructions into the training dataset) to induce the LLM to output a string containing malicious Python code when answering matrix-type math problems. This string is extracted by match_answer(), passed into math_equal(), and ultimately executed by eval(), achieving arbitrary code execution (ACE).

• Quelle: ⚠️ https://github.com/zast-ai/vulnerability-reports/blob/main/bytedance/verl_rce.md
• Benutzer: ZAST.AI (UID 87884)
• Einreichung: 02.04.2026 07:08 (vor 21 Tagen)
• Moderieren: 22.04.2026 20:23 (21 days later)
• Status: Akzeptiert
• VulDB Eintrag: 359040 [ByteDance verl bis 0.7.0 prime_math/grader.py math_equal erweiterte Rechte]
• Punkte: 20