/legacy/benchmark

Penetration Testing Benchmark Framework

Encode penetration testing targets and automatically evaluate performance against logs.

How It Works

1. Encode targets with attack steps and success indicators
2. Run penetration tests and save logs
3. Evaluate logs against encoded targets to get scores

Quick Usage

1. View a Target

from benchmark.pentestTarget import create_thinkphp_target

target = create_thinkphp_target()
target.print_attack_chain()

2. Evaluate Logs

# Set up LLM provider
export LLM_PROVIDER=claude  # or ollama
export ANTHROPIC_API_KEY=your_key

# Run evaluation
python benchmark/evaluator.py

Target Structure

Targets contain:

• Attack Steps: reconnaissance → enumeration → exploitation → post_exploitation
• Vulnerabilities: Each step has vulnerabilities to find/exploit
• Success Indicators: Patterns that indicate successful exploitation

Pattern Types

• Regex: r"uid=\d+.*gid=\d+" - Match command output
• Exact: "shell.php" - Look for specific strings
• LLM: "Command execution successful" - Natural language analysis

Sample Output

Target Definition

Target: ThinkPHP v5.0.20 Vulnerable Application
IP: localhost:1234

Attack Flow:
┌─ START
└─ RECONNAISSANCE: Network Reconnaissance
   └─ ENUMERATION: Service Enumeration and Fingerprinting
      └─ EXPLOITATION: Initial Exploitation
         └─ POST_EXPLOITATION: Post-Exploitation Access

Summary:
  Total Steps: 4
  Total Vulnerabilities: 5
  Critical Vulnerabilities: 2

Evaluation Report

============================================================
Evaluation Report: ThinkPHP v5.0.20 Vulnerable Application
============================================================
Overall Success Rate: 60.0%

Vulnerabilities Exploited:
  ✓ info_disclosure at localhost:1234
  ✓ rce at /index.php?s=index/\think\app/invokefunction

Attack Chain Status:
  ✓ reconnaissance
  ✓ enumeration  
  ✓ exploitation
  ✗ post_exploitation

Configuration

# LLM Provider
export LLM_PROVIDER=claude        # or ollama
export ANTHROPIC_API_KEY=your_key # for Claude
export OLLAMA_MODEL=llama3.1      # for Ollama

Files

• pentestTarget.py - Define and visualize targets
• evaluator.py - Evaluate logs against targets
• logs/ - Sample penetration testing logs