Advisory

PYSEC-2026-2506

Also known as: CVE-2026-44972, GHSA-m5p4-gvpx-4mvr

Summary

GuardDog: Unsanitized human-readable scan output allows terminal escape injection from malicious package content

Details

Summary
GuardDog includes attacker-controlled filenames, file locations, messages, and code snippets in its default human-readable output without escaping terminal control characters. A malicious package can therefore inject ANSI or OSC escape sequences into analyst terminals or CI logs.

Description
The finding formatter stores file paths and snippets from scanned content:

location = file_path + ":" + str(start_line)
finding = {
    "location": location,
    "code": code,
    "message": result["extra"]["message"],
}

The human-readable reporter later prints these values directly:

"  * " + finding["message"] + " at " + finding["location"] + "\n    " + _format_code_line_for_output(finding["code"])

No escaping is applied for control characters such as \x1b. A malicious package can therefore ship a filename like:

evil\x1b[2J.py

or matched source lines containing terminal escapes, which survive into the final CLI output.

Reproduction summary

  1. Create a file whose name contains \x1b[2J.
  2. Feed a semgrep-style result referencing that file into Analyzer._format_semgrep_response().
  3. Render the result with HumanReadableReporter.print_scan_results().
  4. The output string contains the raw escape bytes, which a terminal may interpret.

Key code paths

  • guarddog/analyzer/analyzer.py:377-392
  • guarddog/reporters/human_readable.py:36-42
  • guarddog/reporters/human_readable.py:84-91

Practical impact
This can be used to:

  • clear or rewrite analyst terminal output
  • inject misleading or spoofed log content in CI
  • emit clickable OSC 8 hyperlinks or title changes in compatible terminals

Prior public disclosure check
As of 2026-03-18, no matching public GitHub advisory, CVE, or public repo issue was found for this specific bug.

Suggested fix
Escape or strip terminal control characters before rendering any attacker-controlled value in human-readable output. This should cover package names, file paths, messages, and code snippets.

Severity

Medium Basis: CVSS v3 score 5, vector CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:N

Dates

Published
2026-07-13 15:15 UTC
Modified
2026-07-13 16:31 UTC

Affected packages and fixes

guarddog

PyPI · All systems

Affected

  • introduced 2.6.0, last affected 2.9.0 (ecosystem)

Fixed versions

None listed

Fix

No fixed version is listed.

Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.

Check your own computer: guard packages check

View this record on OSV.dev

Source

Source: PyPA Advisory Database · CC BY 4.0

Sources & licences