from cedric@lemmy.ml to security@lemmy.ml on 01 Jul 19:41
https://lemmy.ml/post/32550706
New AI Model for Chinese Vulnerability Severity in Vulnerability-Lookup!
We just rolled out a new NLP model that estimates the severity of CNVD vulnerabilities using only their Chinese-language descriptions—no CVSS score required.
🔍 Built on top of our Chinese MacBERT-based classifier, this model outputs one of three severity levels: 低 (Low), 中 (Medium), 高 (High)—plus a confidence score.
Why does this matter? CNVD entries often lack standardized CVSS data. With this model, we bridge that gap by using purely textual analysis to provide a severity estimate—making CNVD data far more actionable.
💡 How it works:
- Fine-tuned on our CIRCL/Vulnerability-CNVD dataset
- Served through our in-house ML-Gateway, a FastAPI-based model hub
- Integrated directly into the Vulnerability-Lookup platform
- Fully open-source and privacy-respecting (no external calls)
Our mission is simple: make vulnerability data smarter and more usable—even when structured metadata is missing.
More information about our AI datasets and models:
threaded - newest