CVE-2026-13502: ANTLR4 Maven Plugin TOCTOU Vulnerability Guide

Technical summary

CVE-2026-13502 is a time-of-check time-of-use (TOCTOU) vulnerability in the GrammarDependencies.java component of the ANTLR4 Maven plugin. The flaw resides in the ObjectInputStream.readObject() method, which performs unsafe deserialization without proper synchronization between permission validation and actual file access. An attacker with local system access and sufficient privileges can create a race condition to substitute files between the security check and their actual use, potentially gaining read or write access to sensitive build-time data. The vulnerability requires high attack complexity and low privileges.

Business impact

Organizations using ANTLR4 Maven plugin in their build pipelines face risk of build artifact manipulation and information disclosure. An insider or compromised local account could poison generated code or extract intellectual property from grammar definitions. The impact is primarily confined to development environments rather than production systems, but compromised build artifacts could propagate downstream. The practical risk depends on build system isolation and the sensitivity of grammar and generated code within your organization.

Affected systems

ANTLR4 versions up to and including 4.13.2 are affected, specifically the Maven plugin component. No vendor product list was provided in the vulnerability data. Organizations should audit their development and CI/CD infrastructure for Maven plugin usage, particularly build servers and developer workstations where the plugin executes.

Exploitability

While a proof-of-concept has been published, real-world exploitation remains difficult. The attack vector is strictly local; network-based exploitation is not possible. An attacker must either have direct system access or compromise an existing account with plugin execution privileges. The race condition window is narrow and requires precise timing, making this not a trivial "one-click" exploit. The vendor's non-response suggests no active in-the-wild exploitation campaigns have been observed, but the published proof-of-concept raises awareness among determined threat actors.

Remediation

Users of ANTLR4 should update the Maven plugin beyond version 4.13.2. Verify the exact patched version against the official ANTLR project releases. Organizations should also review build system access controls: restrict who can execute Maven builds, isolate build servers, and use immutable artifact repositories where possible to detect tampering.

Patch guidance

Check the official ANTLR4 GitHub repository and Maven Central for versions released after 4.13.2. Apply the update to all development environments, CI/CD systems, and build servers where the Maven plugin is configured. Coordinate the update with your build pipeline validation to ensure compatibility with your grammar definitions. Test the patched version in a staging environment before production deployment.

Detection guidance

Monitor local file systems on build servers for unexpected file operations during Maven plugin execution. Watch for TOCTOU-style activity patterns: rapid file permission checks followed immediately by file operations on the same objects. Review audit logs for GrammarDependencies class instantiation and ObjectInputStream deserialization events. Inspect Maven build logs for anomalies in plugin initialization. Consider integrity monitoring on grammar files and generated source code directories to catch unauthorized modifications post-build.

Why prioritize this

This vulnerability merits prompt but not emergency attention. The MEDIUM severity reflects local-only attack vector and high complexity requirements. However, the published proof-of-concept and vendor non-response eliminate the 'unknown' factor. Prioritize patching if your organization heavily relies on ANTLR for code generation in sensitive domains (compilers, DSLs for financial or security-critical systems). If ANTLR usage is light or confined to non-sensitive projects, patch within normal maintenance windows.

Risk score, explained

The CVSS 4.5 score reflects a constrained threat model: local-only access (AV:L), high attack complexity due to race condition timing (AC:H), low privileges sufficient but not zero (PR:L), no user interaction required (UI:N), and confidentiality, integrity, and availability impact all possible but limited (C:L/I:L/A:L). The score appropriately penalizes this as a targeted vulnerability relevant primarily to insider threats and compromised developer accounts rather than a mass-exploitation vector.

Frequently asked questions