| --- |
| base_model: meta-llama/Meta-Llama-3.1-8B-Instruct |
| language: |
| - en |
| license: mit |
| tags: |
| - lora |
| - security |
| - cybersecurity |
| - peft |
| --- |
| |
| # Security Testing Agent — LoRA |
|
|
| Defensive security testing assistant. Analyzes code for vulnerabilities and suggests fixes. |
|
|
| **Base model:** meta-llama/Meta-Llama-3.1-8B-Instruct |
|
|
| ## Load it |
|
|
| ```python |
| from peft import PeftModel |
| from transformers import AutoModelForCausalLM, AutoTokenizer |
| |
| model = AutoModelForCausalLM.from_pretrained("meta-llama/Meta-Llama-3.1-8B-Instruct") |
| model = PeftModel.from_pretrained(model, "Xx-Vexento-xX/security-testing-agent-lora") |
| tokenizer = AutoTokenizer.from_pretrained("Xx-Vexento-xX/security-testing-agent-lora") |
| ``` |
|
|
| ## Training |
| - 15 curriculum rounds across Python, JavaScript, Java, Go, PHP, Ruby, TypeScript, Rust, Kotlin, C# |
| - Vulnerability types: SQL Injection, XSS, CSRF, SSRF, Command Injection, IDOR, XXE, JWT weaknesses, Path Traversal, Prototype Pollution, and more |
| - RAG-aware: recognizes [REFERENCE CONTEXT] blocks for external knowledge injection |
|
|
| ## Intended use |
| Authorized defensive security testing of your own code only. |
| Not for unauthorized access or offensive purposes. |
