AI-Powered Reconnaissance

#!/usr/bin/env python3
"""Post-process BBOT scan results with LLM analysis."""

import json
import subprocess
from pathlib import Path
from openai import OpenAI

client = OpenAI()

def load_bbot_results(scan_dir: str) -> dict:
    """Load BBOT output events from JSON lines file."""
    events = []
    output_file = Path(scan_dir) / "output.ndjson"
    
    with open(output_file) as f:
        for line in f:
            event = json.loads(line.strip())
            events.append(event)
    
    # Categorize events
    subdomains = [e for e in events if e.get("type") == "DNS_NAME"]
    open_ports = [e for e in events if e.get("type") == "OPEN_TCP_PORT"]
    urls = [e for e in events if e.get("type") == "URL"]
    techs = [e for e in events if e.get("type") == "TECHNOLOGY"]
    findings = [e for e in events if e.get("type") == "FINDING"]
    
    return {
        "subdomains": subdomains,
        "open_ports": open_ports,
        "urls": urls,
        "technologies": techs,
        "findings": findings
    }

def analyze_with_llm(results: dict, target: str) -> str:
    """Send BBOT results to LLM for analysis and prioritization."""
    
    summary = f"""Target: {target}
Subdomains found: {len(results['subdomains'])}
Open ports: {len(results['open_ports'])}
URLs discovered: {len(results['urls'])}
Technologies detected: {len(results['technologies'])}
Findings: {len(results['findings'])}

Key subdomains:
{json.dumps([s['data'] for s in results['subdomains'][:30]], indent=2)}

Open ports:
{json.dumps([p['data'] for p in results['open_ports'][:30]], indent=2)}

Technologies:
{json.dumps([t['data'] for t in results['technologies'][:20]], indent=2)}

Findings:
{json.dumps([f['data'] for f in results['findings'][:15]], indent=2)}"""

    response = client.chat.completions.create(
        model="gpt-5",
        messages=[
            {"role": "system", "content": """You are an expert penetration tester 
analyzing reconnaissance results. Identify:
1. High-value targets (admin panels, APIs, staging envs)
2. Potential attack vectors based on tech stack
3. Misconfigurations or exposed services
4. Prioritized next steps for exploitation
Be specific and actionable. Reference actual findings."""},
            {"role": "user", "content": summary}
        ],
        temperature=0.3
    )
    return response.choices[0].message.content

# Usage
results = load_bbot_results("/tmp/bbot-full/scan_name")
analysis = analyze_with_llm(results, "example.com")
print(analysis)

#!/usr/bin/env python3
"""Subfinder + AI enrichment pipeline."""

import subprocess
import json
from openai import OpenAI

client = OpenAI()

def run_subfinder(domain: str) -> list[str]:
    """Run Subfinder and return discovered subdomains."""
    result = subprocess.run(
        ["subfinder", "-d", domain, "-silent", "-all"],
        capture_output=True, text=True, timeout=300
    )
    return [s.strip() for s in result.stdout.strip().split("\n") if s.strip()]

def enrich_with_httpx(subdomains: list[str]) -> list[dict]:
    """Probe subdomains with httpx for live hosts and tech."""
    input_data = "\n".join(subdomains)
    result = subprocess.run(
        ["httpx", "-json", "-silent",
         "-status-code", "-title", "-tech-detect",
         "-follow-redirects", "-timeout", "10"],
        input=input_data,
        capture_output=True, text=True, timeout=600
    )
    hosts = []
    for line in result.stdout.strip().split("\n"):
        if line.strip():
            hosts.append(json.loads(line))
    return hosts

def ai_prioritize(domain: str, hosts: list[dict]) -> str:
    """Use LLM to prioritize discovered hosts."""
    host_summary = json.dumps(hosts[:50], indent=2)
    
    response = client.chat.completions.create(
        model="gpt-5",
        messages=[
            {"role": "system", "content": """Analyze these subdomain results 
for a penetration test. Classify each into:
- CRITICAL: Admin panels, staging, internal tools, APIs
- HIGH: Login pages, file uploads, old software versions
- MEDIUM: Standard web apps with potential attack surface
- LOW: Static content, CDN, marketing pages
Explain your reasoning and suggest exploitation approaches."""},
            {"role": "user", "content": f"Domain: {domain}\n\nHosts:\n{host_summary}"}
        ],
        temperature=0.2
    )
    return response.choices[0].message.content

# Pipeline execution
domain = "example.com"
subs = run_subfinder(domain)
print(f"[*] Found {len(subs)} subdomains")

live_hosts = enrich_with_httpx(subs)
print(f"[*] {len(live_hosts)} live hosts detected")

analysis = ai_prioritize(domain, live_hosts)
print(analysis)

# Install Katana (CGO enabled for headless/parser features)
CGO_ENABLED=1 go install github.com/projectdiscovery/katana/cmd/katana@latest

# Basic crawl
katana -u https://example.com -o crawl.txt

# Deep crawl with JavaScript rendering
katana -u https://example.com \
  -js-crawl \
  -headless \
  -depth 5 \
  -pls heuristic \
  -dwt 3 \
  -known-files all \
  -o deep-crawl.txt

# Extract endpoints and parameters
katana -u https://example.com \
  -js-crawl \
  -headless \
  -field url,path,fqdn \
  -o endpoints.txt

# Crawl multiple targets from Subfinder
subfinder -d example.com -silent | \
  httpx -silent | \
  katana -js-crawl -headless -depth 3 -o full-crawl.txt

# Output as JSONL with all fields
katana -u https://example.com \
  -jsonl \
  -js-crawl \
  -headless \
  -field-config /path/to/field-config.yaml \
  -o crawl.json

# Crawl while excluding static assets, then grep output for secret-like strings
katana -u https://example.com \
  -js-crawl \
  -headless \
  -ef png,jpg,gif,svg,css,woff,ttf,woff2,eot \
  | grep -iE "(api[_-]?key|secret|token|password|auth)"

#!/usr/bin/env python3
"""AI-Powered OSINT Pipeline for Reconnaissance."""

import json
import subprocess
import re
from pathlib import Path
from dataclasses import dataclass, asdict
from openai import OpenAI

client = OpenAI()

@dataclass
class OSINTResult:
    source: str
    data_type: str
    content: str
    confidence: float = 0.0
    risk_score: float = 0.0

class AIOSINTPipeline:
    def __init__(self, target_domain: str):
        self.target = target_domain
        self.results: list[OSINTResult] = []
    
    def dns_recon(self) -> list[OSINTResult]:
        """Enumerate DNS records and analyze with AI."""
        records = []
        for rtype in ["A", "AAAA", "MX", "NS", "TXT", "CNAME", "SOA"]:
            result = subprocess.run(
                ["dig", "+short", self.target, rtype],
                capture_output=True, text=True, timeout=30
            )
            if result.stdout.strip():
                records.append(OSINTResult(
                    source="DNS",
                    data_type=rtype,
                    content=result.stdout.strip()
                ))
        self.results.extend(records)
        return records
    
    def whois_analysis(self) -> OSINTResult:
        """WHOIS lookup with AI interpretation."""
        result = subprocess.run(
            ["whois", self.target],
            capture_output=True, text=True, timeout=30
        )
        whois_data = OSINTResult(
            source="WHOIS",
            data_type="registration",
            content=result.stdout[:3000]  # Truncate for LLM context
        )
        self.results.append(whois_data)
        return whois_data
    
    def github_recon(self) -> list[OSINTResult]:
        """Search GitHub for exposed secrets and architecture clues."""
        search_terms = [
            f'"{self.target}" password',
            f'"{self.target}" api_key OR secret_key',
            f'"{self.target}" internal OR staging',
            f'org:{self.target.split(".")[0]} filename:.env',
        ]
        findings = []
        for term in search_terms:
            result = subprocess.run(
                ["gh", "search", "code", term, "--json",
                 "repository,path,textMatches", "-L", "10"],
                capture_output=True, text=True, timeout=30
            )
            if result.stdout.strip():
                findings.append(OSINTResult(
                    source="GitHub",
                    data_type="code_search",
                    content=result.stdout[:2000]
                ))
        self.results.extend(findings)
        return findings
    
    def certificate_transparency(self) -> list[OSINTResult]:
        """Query Certificate Transparency logs."""
        result = subprocess.run(
            ["curl", "-s",
             f"https://crt.sh/?q=%.{self.target}&output=json"],
            capture_output=True, text=True, timeout=30
        )
        if result.stdout.strip():
            try:
                certs = json.loads(result.stdout)
                unique_names = list(set(
                    c.get("name_value", "") for c in certs
                ))
                ct_result = OSINTResult(
                    source="CertTransparency",
                    data_type="certificates",
                    content=json.dumps(unique_names[:100])
                )
                self.results.append(ct_result)
                return [ct_result]
            except json.JSONDecodeError:
                pass
        return []
    
    def ai_analyze_all(self) -> str:
        """Send all collected OSINT to LLM for analysis."""
        results_summary = json.dumps(
            [asdict(r) for r in self.results], indent=2
        )
        
        response = client.chat.completions.create(
            model="gpt-5",
            messages=[
                {"role": "system", "content": """You are an expert OSINT 
analyst performing reconnaissance for an authorized penetration test. 
Analyze all collected data and produce:

1. ORGANIZATION PROFILE: Key facts about the target
2. ATTACK SURFACE: Identified entry points and services
3. EXPOSED SECRETS: Any credentials, keys, or sensitive data found
4. SOCIAL ENGINEERING VECTORS: Information useful for phishing/SE
5. RISK ASSESSMENT: Overall exposure rating with justification
6. RECOMMENDED NEXT STEPS: Prioritized actions for the engagement

Be thorough and reference specific findings."""},
                {"role": "user", "content": f"""Target: {self.target}
                
OSINT Collection Results:
{results_summary}"""}
            ],
            temperature=0.2
        )
        return response.choices[0].message.content

# Execute pipeline
pipeline = AIOSINTPipeline("example.com")
pipeline.dns_recon()
pipeline.whois_analysis()
pipeline.github_recon()
pipeline.certificate_transparency()
report = pipeline.ai_analyze_all()
print(report)

#!/usr/bin/env python3
"""Complete AI Recon Pipeline — chains multiple tools with LLM analysis."""

import json
import subprocess
import asyncio
from datetime import datetime
from pathlib import Path
from dataclasses import dataclass, field, asdict
from openai import OpenAI

client = OpenAI()

@dataclass
class ReconReport:
    target: str
    timestamp: str = field(
        default_factory=lambda: datetime.now().isoformat()
    )
    subdomains: list[str] = field(default_factory=list)
    live_hosts: list[dict] = field(default_factory=list)
    open_ports: list[dict] = field(default_factory=list)
    technologies: list[dict] = field(default_factory=list)
    vulnerabilities: list[dict] = field(default_factory=list)
    ai_analysis: str = ""

def run_cmd(cmd: list[str], timeout: int = 300) -> str:
    """Run a command and return stdout."""
    result = subprocess.run(
        cmd, capture_output=True, text=True, timeout=timeout
    )
    return result.stdout.strip()

def stage_1_subdomain_enum(target: str) -> list[str]:
    """Stage 1: Subdomain enumeration with multiple tools."""
    print(f"[1/5] Subdomain enumeration for {target}")
    
    # Subfinder
    sf_output = run_cmd(["subfinder", "-d", target, "-silent", "-all"])
    subs = set(sf_output.split("\n")) if sf_output else set()
    
    # Amass (passive)
    amass_output = run_cmd(
        ["amass", "enum", "-passive", "-d", target],
        timeout=600
    )
    if amass_output:
        subs.update(amass_output.split("\n"))
    
    # Certificate Transparency
    ct_output = run_cmd([
        "curl", "-s",
        f"https://crt.sh/?q=%.{target}&output=json"
    ])
    if ct_output:
        try:
            certs = json.loads(ct_output)
            for cert in certs:
                name = cert.get("name_value", "")
                for sub in name.split("\n"):
                    if sub.strip().endswith(target):
                        subs.add(sub.strip())
        except json.JSONDecodeError:
            pass
    
    return sorted(list(subs))

def stage_2_probe_hosts(subdomains: list[str]) -> list[dict]:
    """Stage 2: Probe for live hosts with httpx."""
    print(f"[2/5] Probing {len(subdomains)} subdomains")
    
    input_data = "\n".join(subdomains)
    result = subprocess.run(
        ["httpx", "-json", "-silent",
         "-status-code", "-title", "-tech-detect",
         "-content-length", "-follow-redirects"],
        input=input_data,
        capture_output=True, text=True, timeout=600
    )
    
    hosts = []
    for line in result.stdout.strip().split("\n"):
        if line.strip():
            try:
                hosts.append(json.loads(line))
            except json.JSONDecodeError:
                continue
    return hosts

def stage_3_port_scan(subdomains: list[str]) -> list[dict]:
    """Stage 3: Port scanning with Naabu."""
    print(f"[3/5] Port scanning top targets")
    
    # Scan top 1000 ports on live hosts
    input_data = "\n".join(subdomains[:50])  # Limit scope
    result = subprocess.run(
        ["naabu", "-json", "-silent", "-top-ports", "1000"],
        input=input_data,
        capture_output=True, text=True, timeout=600
    )
    
    ports = []
    for line in result.stdout.strip().split("\n"):
        if line.strip():
            try:
                ports.append(json.loads(line))
            except json.JSONDecodeError:
                continue
    return ports

def stage_4_vuln_scan(live_hosts: list[dict]) -> list[dict]:
    """Stage 4: Vulnerability scanning with Nuclei."""
    print(f"[4/5] Vulnerability scanning {len(live_hosts)} hosts")
    
    urls = [h.get("url", "") for h in live_hosts if h.get("url")]
    input_data = "\n".join(urls)
    
    result = subprocess.run(
        ["nuclei", "-json", "-silent",
         "-severity", "medium,high,critical",
         "-rate-limit", "50"],
        input=input_data,
        capture_output=True, text=True, timeout=900
    )
    
    vulns = []
    for line in result.stdout.strip().split("\n"):
        if line.strip():
            try:
                vulns.append(json.loads(line))
            except json.JSONDecodeError:
                continue
    return vulns

def stage_5_ai_analysis(report: ReconReport) -> str:
    """Stage 5: Comprehensive AI analysis of all findings."""
    print("[5/5] AI analysis of results")
    
    context = f"""Target: {report.target}
Scan timestamp: {report.timestamp}

SUBDOMAINS ({len(report.subdomains)} total):
{json.dumps(report.subdomains[:40], indent=2)}

LIVE HOSTS ({len(report.live_hosts)} total):
{json.dumps(report.live_hosts[:30], indent=2)}

OPEN PORTS ({len(report.open_ports)} total):
{json.dumps(report.open_ports[:30], indent=2)}

TECHNOLOGIES:
{json.dumps(report.technologies[:20], indent=2)}

VULNERABILITIES ({len(report.vulnerabilities)} total):
{json.dumps(report.vulnerabilities[:20], indent=2)}"""

    response = client.chat.completions.create(
        model="gpt-5",
        messages=[
            {"role": "system", "content": """You are a senior penetration 
tester analyzing automated reconnaissance results. Produce a structured 
report with:

## Executive Summary
Brief overview of the target attack surface.

## Critical Findings
High-risk items requiring immediate attention.

## Attack Vectors
Specific, actionable exploitation paths based on findings.

## Technology Stack Analysis
Identified technologies and their known vulnerabilities.

## Recommended Exploitation Order
Prioritized list of targets to exploit, with reasoning.

## OPSEC Considerations
Detection risks and how to avoid them.

Be precise. Reference specific subdomains, ports, and vulnerabilities."""},
            {"role": "user", "content": context}
        ],
        temperature=0.2,
        max_tokens=4000
    )
    return response.choices[0].message.content

def run_pipeline(target: str) -> ReconReport:
    """Execute the full recon pipeline."""
    report = ReconReport(target=target)
    
    # Stage 1: Subdomain enumeration
    report.subdomains = stage_1_subdomain_enum(target)
    print(f"    Found {len(report.subdomains)} subdomains")
    
    # Stage 2: Live host detection
    report.live_hosts = stage_2_probe_hosts(report.subdomains)
    print(f"    {len(report.live_hosts)} live hosts")
    
    # Stage 3: Port scanning
    report.open_ports = stage_3_port_scan(report.subdomains)
    print(f"    {len(report.open_ports)} open ports")
    
    # Stage 4: Vulnerability scanning
    report.vulnerabilities = stage_4_vuln_scan(report.live_hosts)
    print(f"    {len(report.vulnerabilities)} vulnerabilities")
    
    # Stage 5: AI analysis
    report.ai_analysis = stage_5_ai_analysis(report)
    
    # Save structured report
    output_path = Path(f"recon_{target}_{report.timestamp[:10]}.json")
    with open(output_path, "w") as f:
        json.dump(asdict(report), f, indent=2)
    print(f"\n[*] Report saved to {output_path}")
    
    return report

# Execute
report = run_pipeline("example.com")
print("\n" + report.ai_analysis)

#!/usr/bin/env python3
"""AI-Powered Attack Surface Management — Continuous Monitoring."""

import json
import hashlib
import subprocess
from datetime import datetime
from pathlib import Path
from openai import OpenAI

client = OpenAI()

class AttackSurfaceMonitor:
    def __init__(self, target: str, state_dir: str = "./asm_state"):
        self.target = target
        self.state_dir = Path(state_dir)
        self.state_dir.mkdir(exist_ok=True)
        self.state_file = self.state_dir / f"{target}.json"
    
    def load_previous_state(self) -> dict:
        """Load the previous scan state for comparison."""
        if self.state_file.exists():
            return json.loads(self.state_file.read_text())
        return {"subdomains": [], "hosts": [], "ports": [], "timestamp": ""}
    
    def save_state(self, state: dict):
        """Save current scan state."""
        state["timestamp"] = datetime.now().isoformat()
        self.state_file.write_text(json.dumps(state, indent=2))
    
    def discover_assets(self) -> dict:
        """Run discovery tools and return current state."""
        # Subdomain enumeration
        sf = subprocess.run(
            ["subfinder", "-d", self.target, "-silent", "-all"],
            capture_output=True, text=True, timeout=300
        )
        subdomains = sorted(set(
            s.strip() for s in sf.stdout.split("\n") if s.strip()
        ))
        
        # Live host probing
        input_data = "\n".join(subdomains)
        httpx_out = subprocess.run(
            ["httpx", "-json", "-silent", "-status-code", "-title",
             "-tech-detect", "-content-length"],
            input=input_data,
            capture_output=True, text=True, timeout=600
        )
        hosts = []
        for line in httpx_out.stdout.strip().split("\n"):
            if line.strip():
                try:
                    hosts.append(json.loads(line))
                except json.JSONDecodeError:
                    continue
        
        return {
            "subdomains": subdomains,
            "hosts": hosts,
            "ports": [],  # Add naabu scan if needed
        }
    
    def detect_changes(self, previous: dict, current: dict) -> dict:
        """Compare states and identify changes."""
        prev_subs = set(previous.get("subdomains", []))
        curr_subs = set(current.get("subdomains", []))
        
        prev_urls = set(
            h.get("url", "") for h in previous.get("hosts", [])
        )
        curr_urls = set(
            h.get("url", "") for h in current.get("hosts", [])
        )
        
        return {
            "new_subdomains": sorted(curr_subs - prev_subs),
            "removed_subdomains": sorted(prev_subs - curr_subs),
            "new_hosts": sorted(curr_urls - prev_urls),
            "removed_hosts": sorted(prev_urls - curr_urls),
        }
    
    def ai_risk_assessment(self, changes: dict, current: dict) -> str:
        """Use AI to assess risk of detected changes."""
        if not any(changes.values()):
            return "No changes detected since last scan."
        
        response = client.chat.completions.create(
            model="gpt-5",
            messages=[
                {"role": "system", "content": """You are an attack surface 
management analyst. Assess the security impact of infrastructure changes.
Rate each change: CRITICAL / HIGH / MEDIUM / LOW / INFO.
Recommend immediate actions for high-risk changes."""},
                {"role": "user", "content": f"""Target: {self.target}

Changes detected:
{json.dumps(changes, indent=2)}

Current state summary:
- Total subdomains: {len(current['subdomains'])}
- Live hosts: {len(current['hosts'])}"""}
            ],
            temperature=0.2
        )
        return response.choices[0].message.content
    
    def run(self):
        """Execute monitoring cycle."""
        print(f"[*] ASM scan for {self.target}")
        previous = self.load_previous_state()
        current = self.discover_assets()
        changes = self.detect_changes(previous, current)
        
        # AI risk assessment
        risk_report = self.ai_risk_assessment(changes, current)
        
        # Save new state
        self.save_state(current)
        
        print(f"[+] Subdomains: {len(current['subdomains'])}")
        print(f"[+] Live hosts: {len(current['hosts'])}")
        print(f"[+] New subdomains: {len(changes['new_subdomains'])}")
        print(f"[+] Removed subdomains: {len(changes['removed_subdomains'])}")
        print(f"\n{risk_report}")

# Usage — run on a cron schedule
monitor = AttackSurfaceMonitor("example.com")
monitor.run()