elsiddik/finsec_detector
elsiddik/finsec_detector is a 7.6 billion parameter specialized cybersecurity Large Language Model, based on Qwen2.5-Coder-7B-Instruct and fine-tuned with 4-bit QLoRA. It functions as an automated Senior Application Security Auditor, capable of auditing source code across 9 programming languages. The model identifies vulnerabilities, classifies them by severity and CWE ID, and generates ready-to-merge secure code patches in structured JSON format, achieving 100% precision and 83.3% recall on multi-language vulnerability benchmarks.
Loading preview...
FineSec-Detector: Specialized Security LLM
elsiddik/finsec_detector is a 7.6 billion parameter Large Language Model (LLM) specifically designed for cybersecurity applications. Built upon the Qwen2.5-Coder-7B-Instruct architecture and fine-tuned using 4-bit QLoRA, this model excels as an automated Senior Application Security (AppSec) Auditor.
Key Capabilities
- Automated Vulnerability Detection: Audits source code in Python, C/C++, JavaScript, Go, PHP, Java, and Bash.
- Structured Output: Generates security reports in standardized JSON format, suitable for integration into CI/CD pipelines.
- Classification: Identifies and classifies vulnerabilities into standard CWE categories (e.g., CWE-89 SQLi, CWE-79 XSS) with CVSS-aligned severity levels (CRITICAL, HIGH, MEDIUM, LOW).
- Remediation: Produces secure code refactors and diffs to directly patch vulnerable logic.
Performance Highlights
Evaluated on multi-language vulnerability benchmarks, FineSec-Detector demonstrates exceptional accuracy:
- Precision Rate: 100.0% (zero false positives)
- Detection Recall: 83.3% across various languages
- F1 Rating Score: 90.9% (outstanding overall balance)
Good For
- Automating security audits in software development.
- Integrating vulnerability detection and patching into CI/CD workflows.
- Developers and security teams needing precise, actionable security insights and code remediation.