Predict the risk score (CVSS) of a cybersecurity vulnerability (CVE) based on its description using Natural Learning Processing (NLP)
The flow of a CVE analysis starts with detecting it and writing a description by researcher 1, and then researcher 2 computes a score based on CVSS guidelines. Our model goal is to replace researcher 2 and allow fast alerting of new vulnerabilities that require attention.
This code is apart of a Kaggle competition Vulnerability risk scoring In this project, we use tfidf with different ensemble techniques: Random Forest, CatBoost, AdaBoost and Stacking.
The notebook present analysis of the problem and exploratory data analysis.
References:
- CVE - cyber vulnerability: https://cve.mitre.org/
- CSS Score : https://nvd.nist.gov/vuln-metrics/cvss
- Data Source: https://github.com/cornerpirate/cve-offline