Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

When "Correct" Is Not Safe: Can We Trust Functionally Correct Patches Generated by Code Agents?

Yibo Peng, James Song, Lei Li, Xinyu Yang, Mihai Christodorescu, Ravi Mangal, Corina Pasareanu, Haizhong Zheng, Beidi Chen

TL;DR

This paper reveals a novel type of threat to real-world code agents: Functionally Correct yet Vulnerable (FCV) patches, which pass all test cases but contain vulnerable code.

Abstract

Code agents are increasingly trusted to autonomously fix bugs on platforms such as GitHub, yet their security evaluation focuses almost exclusively on functional correctness. In this paper, we reveal a novel type of threat to real-world code agents: Functionally Correct yet Vulnerable (FCV) patches, which pass all test cases but contain vulnerable code. With our proposed FCV-Attack, which can be deliberately crafted by malicious attackers or implicitly introduced by benign developers, we show that SOTA LLMs (e.g., ChatGPT and Claude) and agent scaffolds (e.g., SWE-agent and OpenHands) are all vulnerable to this FCV threat; across 12 agent-model combinations on SWE-Bench, the attack only requires black-box access and a single query to the code agent to perform the attack. For example, for CWE-538 (information exposure vulnerability), the FCV-Attack attains an attack success rate of $40.7\%$ on GPT-5 Mini + OpenHands. Our results reveal an important security threat overlooked by current evaluation paradigms and urge the development of security-aware defenses for code agents.

When "Correct" Is Not Safe: Can We Trust Functionally Correct Patches Generated by Code Agents?

TL;DR

This paper reveals a novel type of threat to real-world code agents: Functionally Correct yet Vulnerable (FCV) patches, which pass all test cases but contain vulnerable code.

Abstract

Code agents are increasingly trusted to autonomously fix bugs on platforms such as GitHub, yet their security evaluation focuses almost exclusively on functional correctness. In this paper, we reveal a novel type of threat to real-world code agents: Functionally Correct yet Vulnerable (FCV) patches, which pass all test cases but contain vulnerable code. With our proposed FCV-Attack, which can be deliberately crafted by malicious attackers or implicitly introduced by benign developers, we show that SOTA LLMs (e.g., ChatGPT and Claude) and agent scaffolds (e.g., SWE-agent and OpenHands) are all vulnerable to this FCV threat; across 12 agent-model combinations on SWE-Bench, the attack only requires black-box access and a single query to the code agent to perform the attack. For example, for CWE-538 (information exposure vulnerability), the FCV-Attack attains an attack success rate of on GPT-5 Mini + OpenHands. Our results reveal an important security threat overlooked by current evaluation paradigms and urge the development of security-aware defenses for code agents.
Paper Structure (58 sections, 12 figures, 5 tables)

This paper contains 58 sections, 12 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Revisiting the assumption that correctness implies security.
  3. A practical black-box FCV attack and large-scale evaluation.
  4. Internal State Contamination.
  5. Related Work
  6. Code Agents for Software Engineering.
  7. Security of LLM-Generated Code.
  8. Adversarial Attacks on LLM Agents.
  9. Why "Correct" Is Not Secure: Motivating Observation
  10. Functionally Correct yet Vulnerable (FCV).
  11. Quantifying FCV Risk.
  12. Amplifying Vulnerabilities with FCV-Attack
  13. Threat Model
  14. Attacker Goal
  15. Attacker Capabilities.
  16. ...and 43 more sections

Figures (12)

  • Figure 1: An FCV attack can be initiated through two real-world pathways: deliberately by a malicious contributor, or more subtly, when a benign developer inadvertently incorporates content from a contaminated source. Both pathways result in the same input of developer-style instructions within an issue description, making them indistinguishable from the agent's perspective. This illustrates the core of the FCV threat: functionally correct patches that pass all tests can still embed exploitable vulnerabilities (e.g., CWE-94).
  • Figure 2: Vulnerability rates among functionally correct patches under clean settings.
  • Figure 3: Conceptual examples of Functionally Correct but Vulnerable (FCV) patches. Each patch is designed to resolve a functional issue and pass corresponding tests, yet stealthily embeds a distinct security vulnerability.
  • Figure 4: Average (ASR) across (a) agents, (b) CWE vulnerability types, and (c) LLM models. The results show that SWE-Agent, the CWE-538 attack, and more advanced models like Claude Sonnet 4 yield the highest ASR.
  • Figure 5: FCV rate by number of API calls across four models for Mini-SWE-Agent.
  • ...and 7 more figures