Loading…
ML Artifact & Supply-Chain Integrity
Block Backdoored ML Before It Reaches Prod.
Patent-pending scanning across pickle, .pth, .safetensors, ONNX, PyTorch and TensorFlow checkpoints — supply-chain integrity evidence aligned to ISO 42001 and EU AI Act. Compromised artifacts blocked at the deployment gate.
7
ML Formats Scanned
40+
AI GRC Frameworks
2
Deployment Modes
Pre-Deploy
Release Gate
The Challenge
A Pickle File Can Execute Arbitrary Code on Load.
Models from public hubs, partner-provided checkpoints, and even internally trained artifacts moving between environments carry latent supply-chain risk. Generic software scanners do not understand ML formats. Runtime LLM firewalls cannot detect a backdoor that was already baked into the weights.
No ML Artifact Scanning
Most enterprises deploy imported models, fine-tuned checkpoints, and third-party weights without scanning. The first signal of a compromised artifact is typically a production incident.
Generic Scanners Miss ML Formats
Conventional software composition analysis tooling scans for known CVEs in software dependencies. Pickle payloads, malicious hooks in weights, and ML loader injections are opaque to that class of scanner.
Runtime Firewalls Are Too Late
LLM-firewall tools moderate prompts at runtime. They cannot detect a backdoor already baked into the model — the compromise occurred before the firewall saw the first prompt.
How It Works
Pre-Deployment Release Gate
Step 01
Inspect
Scan pickle, .pth, .safetensors, ONNX, serialized PyTorch and TensorFlow checkpoints, Keras .h5, joblib dumps, and accompanying deployment code.
Step 02
Detect
Identify backdoor payloads in deserialization paths, malicious hooks (exec, subprocess, network call-out, encoded binaries), and dependency injections in loader scripts.
Step 03
Quarantine
Block compromised artifacts at the deployment gate. Quarantine workflow with provenance trace produces full forensic context for the security team.
Step 04
Evidence
Generate PrismMLScanner reports, provenance traces, and gate decision logs pre-mapped to ISO 42001 supply-chain controls and EU AI Act Art. 15.
Key Benefits
Why Teams Choose PrismMLScanner
Core
Comprehensive ML Format Coverage
Pickle, PyTorch (.pth, .pt), .safetensors, ONNX, TensorFlow SavedModel and HDF5, Keras .h5, scikit-learn joblib, plus loader scripts, requirements files, Dockerfiles.
Backdoor Payload Detection
Identifies hidden payloads in pickle deserialization paths, exec/subprocess/network call-out hooks, encoded binary payloads, and malicious dependency injections in deployment code.
Pre-Deployment Release Gate
Integrates into MLOps pipelines as a deployment gate. Compromised artifacts blocked before production with full quarantine and provenance trace.
Patent-Pending Methodology
Built on SISA's patent-pending AI model scanning solution. Detection rules are forensics-informed by SISA's Global PFI breach intelligence.
Standalone or Integrated
Deploy independently as a release gate, or as a release gate before PrismSecure within the full PRISM lifecycle. Artifact integrity feeds PrismGovern.
Audit-Ready Evidence
PrismMLScanner reports, provenance traces, quarantine records, and gate decision logs pre-mapped to ISO 42001, EU AI Act Art. 15, NIST AI RMF MAP/MANAGE, HITRUST AI Security Certification.
Built For Your Team
Designed for Every Stakeholder
CISO
Close the pickle-file and serialized-model attack vector at the gate — not during a forensic investigation.
AI / ML Platform Lead
Imported third-party models become trustable. Internal artifacts moving between environments pass the same gate.
MLOps Lead
Pipeline-integrated release gate with quarantine workflow — supply-chain risk reduced and evidenced at the artifact level.
Compliance / Risk Lead
Audit-ready evidence aligned to ISO 42001 supply-chain controls, EU AI Act Art. 15, NIST AI RMF, HITRUST AI Security Certification.
Formats & Frameworks
Pickle (.pkl)PyTorch (.pth).safetensorsONNXTensorFlow SavedModelKeras .h5joblibHugging Face HubISO 42001EU AI Act Art. 15NIST AI RMFHITRUST AI Security Certification