Regular Expression Denial of Service in Red Hat OpenShift AI

Executive Summary

CVE-2026-15154 is a Regular Expression Denial of Service (ReDoS) vulnerability found in the guardrails-detectors component of Red Hat OpenShift AI. It allows a remote attacker to send specially crafted regular expressions to the public detection API, which are then processed without safeguards. This causes catastrophic backtracking in the regex engine, leading to a worker process consuming 100% CPU indefinitely.

The vulnerability arises because the regex detector does not impose timeouts or complexity limits on the regular expressions it processes. An attacker can exploit this by sending malicious regex patterns that cause the Python regex engine to hang, exhausting system resources.

Useful 0 Not Useful 0

Impact Analysis

This vulnerability can cause a denial of service by making the worker processes consume 100% CPU indefinitely, potentially exhausting system resources.

If multiple concurrent malicious requests are made, they can fully exhaust the detector's resources, which depending on the system configuration, may either cause the entire guardrails-mediated LLM pipeline to become unavailable or allow attackers to bypass the guardrails, gaining unfiltered access to the language model.

Useful 0 Not Useful 0

Detection Guidance

This vulnerability can be detected by monitoring for unusually high CPU usage on the worker processes handling the guardrails-detectors component, especially when processing requests to the public detection API.

Since the vulnerability is triggered by specially crafted regular expressions sent via HTTP requests, inspecting incoming API requests for suspicious or complex regex patterns such as "(a+)+$" can help identify exploitation attempts.

You can use system monitoring commands to detect high CPU usage, for example:

• top or htop to observe CPU usage of worker processes.
• ps aux --sort=-%cpu | head to list processes consuming the most CPU.

To inspect network traffic or logs for suspicious regex patterns in API requests, you might use:

• tcpdump or Wireshark to capture HTTP traffic to the detection API.
• grep or similar tools on API logs to search for regex parameters like "regex" or known malicious patterns.

Useful 0 Not Useful 0

Mitigation Strategies

Immediate mitigation steps include restricting access to the public detection API to trusted sources only, since the attack vector is from an adjacent network.

Implement rate limiting or request throttling on the detection API to reduce the risk of resource exhaustion from multiple concurrent malicious requests.

Monitor and terminate any worker processes consuming excessive CPU to restore service availability.

If possible, apply input validation or filtering on the regex patterns received by the API to block overly complex or potentially malicious regular expressions.

Coordinate with the maintainers or update to a fixed version once available, as the vulnerability stems from lack of safeguards in regex compilation.

Useful 0 Not Useful 0

Compliance Impact

This vulnerability causes a denial of service condition by allowing an attacker to exhaust system resources through a Regular Expression Denial of Service (ReDoS) attack on the guardrails-detectors component. While the CVE description and resources do not explicitly mention compliance with standards such as GDPR or HIPAA, denial of service incidents can impact availability requirements mandated by these regulations.

Specifically, the denial of service could disrupt the availability of the guardrails-mediated LLM pipeline, potentially affecting systems that rely on it for secure or compliant processing. However, there is no direct information indicating data breach, unauthorized access, or data integrity issues that would more directly impact compliance with privacy or security controls under GDPR or HIPAA.

Therefore, the main compliance impact would be related to availability and service continuity obligations, which are part of many regulatory frameworks.

Useful 0 Not Useful 0