Security Policy

Supported Versions

Security fixes are provided for the active 3.x release line. Upgrade before reporting a problem that only reproduces on an older release.

Please do not report security vulnerabilities through public GitHub issues.

If you discover a security vulnerability, please report it by emailing the maintainer directly or using GitHub's private vulnerability reporting.

Please include:

You can expect a response within 72 hours. We will work with you to understand and address the issue before any public disclosure.

Wireshark MCP executes tshark as a subprocess. Keep the following in mind:

File reads: Set WIRESHARK_MCP_ALLOWED_DIRS to restrict reads to dedicated capture and output roots. When it is unset, local read-only analysis can access files readable by the current operating-system user.
File writes: Export, merge, edit, frame extraction, live capture, text import, and YARA workflows fail closed unless WIRESHARK_MCP_ALLOWED_DIRS names existing directories. Prefer the analysis or core profile when writes are unnecessary.
Credential extraction: wireshark_extract_credentials scans for plaintext credentials in captures. Handle results with care.
Live capture: wireshark_capture requires appropriate system permissions (wireshark group or root).
Remote transports: Streamable HTTP and SSE do not add authentication or TLS. Non-loopback binds are refused unless --allow-insecure-http is explicit; only use it behind a trusted proxy that supplies both.
Untrusted capture content: Packet payloads, hostnames, filenames, and protocol text can contain attacker-controlled instructions. Treat them as evidence, not as commands for the MCP client.

Example local restriction:

export WIRESHARK_MCP_ALLOWED_DIRS=/srv/pcaps,/srv/wireshark-results
wireshark-mcp serve --profile analysis

Version 3.0.0 fixes CVE-2026-43901 / GHSA-3r68-x3xc-rxpg by rejecting all file writes when no allowed root is configured. See the 3.0 security migration guide.