Please do not report security vulnerabilities through public GitHub issues.

We take security seriously. If you discover a security vulnerability in MCP Sentinel, please report it responsibly.

Send an email to: security@mcp-sentinel.dev (or create a private security advisory on GitHub)

Include:

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

1. Acknowledgment: Within 48 hours
2. Assessment: Within 7 days
3. Fix & Release: Depends on severity
   • Critical: 1-7 days
   • High: 7-14 days
   • Medium/Low: Next release cycle

• We will notify you when the vulnerability is fixed
• We will credit you in the security advisory (unless you prefer anonymity)
• We request a 90-day embargo before public disclosure

1. Keep MCP Sentinel Updated

   pip install --upgrade mcp-sentinel

2. Verify Downloads

   • Check SHA256 checksums
   • Verify GPG signatures (when available)

3. Run with Least Privilege

   • Don't run as root/administrator
   • Use appropriate file permissions

4. Review Scan Results

   • Act on critical/high findings immediately
   • Don't commit sensitive data revealed by scans

1. Never Commit Secrets

   • Use .gitignore for sensitive files
   • Review diffs before committing
   • Use git-secrets or similar tools

2. Dependency Management

   • Keep dependencies updated
   • Review dependency changes
   • Use pip audit regularly

3. Code Review

   • All code must be reviewed
   • Security-sensitive changes require additional review
   • Use static analysis tools (ruff, mypy)

MCP Sentinel reads files from the filesystem. Ensure you:

• Trust the directories you scan
• Don't scan untrusted symbolic links
• Be aware of file size limits

Some detectors use regex patterns that could be vulnerable to ReDoS (Regular Expression Denial of Service). We:

• Test patterns against ReDoS
• Use timeouts for pattern matching
• Accept PRs that improve pattern safety

We minimize external dependencies and:

• Pin dependency versions in pyproject.toml
• Regularly audit with pip audit
• Review security advisories

Detected secrets are included in scan output with surrounding context. JSON output may contain more detail. Secure your reports appropriately — do not commit SARIF or JSON scan results to public repositories.

1. Secrets: API keys, private keys, credentials (15+ patterns)
2. Code Injection: os.system, subprocess(shell=True), eval, exec, SQL f-strings
3. Prompt Injection: Role manipulation, jailbreaks, system prompt exposure
4. Tool Poisoning: Invisible Unicode, override directives, sensitive path targeting
5. Path Traversal: ../ sequences, zip slip, unsafe open() calls
6. Config Security: DEBUG=True, open CORS, SSL_VERIFY=False, weak secrets
7. SSRF: Unvalidated URL variables in HTTP clients, cloud metadata endpoints
8. Network Binding: Servers bound to 0.0.0.0 instead of 127.0.0.1
9. Missing Auth: Routes and endpoints without authentication
10. Supply Chain: Encoded payloads, install-time exec/network, covert exfiltration
11. Weak Crypto: MD5/SHA-1, insecure random, ECB mode, deprecated ciphers
12. Insecure Deserialization: pickle.loads, yaml.load, ObjectInputStream, PHP unserialize

• No Network Scanning: We don't scan remote servers
• No Code Execution: We only analyze, never execute
• No Data Collection: No telemetry or analytics
• No Cloud Storage: Everything stays local
• No Multi-line Taint: Cross-line variable-to-sink flows require semantic analysis (planned v0.5)

We believe in responsible disclosure. If you discover a vulnerability:

1. Contact us privately first
2. Give us time to fix it
3. Coordinate public disclosure

We will:

• Acknowledge your contribution
• Keep you updated on progress
• Credit you in release notes (unless anonymous preferred)

Contributors who responsibly disclose vulnerabilities:

(None yet - be the first!)

Subscribe to security updates:

• Watch this repository
• Follow releases

• OWASP Top 10
• OWASP Agentic AI Top 10
• Python Security Advisories
• CWE Database

• Security Issues: security@mcp-sentinel.dev
• General Questions: Use GitHub Discussions

Thank you for helping keep MCP Sentinel secure!