Stop Sensitive Data From
Leaking Into AI Tools

# AI DLP Policy Matrix — Data Classification × AI Tool Category
# Actions: ALLOW | LOG | BLOCK | BLOCK+ALERT

policy_matrix:
  "Public":
    text_language:      ALLOW
    code_development:   ALLOW
    image_generation:   ALLOW
    data_analysis:      ALLOW
    voice_audio:        ALLOW

  "Internal":
    text_language:      LOG          # approved tools only
    code_development:   BLOCK        # no internal code in AI assistants
    image_generation:   LOG
    data_analysis:      BLOCK        # internal datasets stay internal
    voice_audio:        LOG

  "Confidential":
    text_language:      BLOCK+ALERT
    code_development:   BLOCK+ALERT
    image_generation:   BLOCK+ALERT
    data_analysis:      BLOCK+ALERT
    voice_audio:        BLOCK+ALERT

  "Restricted":
    # ALL categories blocked. Incident auto-created. Manager notified.
    default_action:     BLOCK+ALERT+INCIDENT

#!/usr/bin/env python3
"""Monitor proxy logs for sensitive data exfiltration to AI tools.
   Reads Squid/BlueCoat-format proxy logs and flags suspicious transfers."""

import re
import csv
import sys
from collections import defaultdict
from datetime import datetime
from urllib.parse import urlparse

SENSITIVE_PATTERNS = {
    "SSN":          re.compile(r"\b\d{3}-\d{2}-\d{4}\b"),
    "CreditCard":   re.compile(r"\b(?:4\d{15}|5[1-5]\d{14}|3[47]\d{13})\b"),
    "AWSKey":       re.compile(r"AKIA[0-9A-Z]{16}"),
    "PrivateKey":   re.compile(r"-----BEGIN (?:RSA )?PRIVATE KEY-----"),
    "APISecret":    re.compile(r"(?:api[_-]?key|api[_-]?secret|token)\s*[:=]\s*['\"][A-Za-z0-9]{20,}", re.I),
    "InternalDoc":  re.compile(r"(?:CONFIDENTIAL|INTERNAL USE ONLY|DO NOT DISTRIBUTE)", re.I),
}

def load_ai_domains(feed_path: str) -> set:
    """Load AI tool domains from the blocklist feed."""
    domains = set()
    with open(feed_path, "r") as f:
        reader = csv.DictReader(f)
        for row in reader:
            domains.add(row["domain"].lower())
    return domains

def analyze_proxy_log(log_path: str, ai_domains: set):
    """Scan proxy log for data exfiltration to AI tool domains."""
    alerts = []
    with open(log_path, "r") as f:
        for line_num, line in enumerate(f, 1):
            parts = line.strip().split()
            if len(parts) < 7: continue

            method = parts[3]
            url = parts[4]
            bytes_sent = int(parts[5]) if parts[5].isdigit() else 0
            domain = urlparse(url).hostname or ""

            if domain.lower() not in ai_domains:
                continue

            alert = {"line": line_num, "domain": domain,
                     "method": method, "bytes": bytes_sent, "flags": []}

            # Flag large POST/PUT as potential data uploads
            if method in ("POST", "PUT") and bytes_sent > 5000:
                alert["flags"].append(f"LARGE_UPLOAD:{bytes_sent}B")
            if method == "POST":
                alert["flags"].append("DATA_SUBMISSION")

            if alert["flags"]:
                alerts.append(alert)

    return alerts

# Run analysis
ai_domains = load_ai_domains("ai_tools_blocklist.csv")
alerts = analyze_proxy_log("proxy_access.log", ai_domains)

print(f"AI Data Exfiltration Report — {datetime.now().strftime('%Y-%m-%d %H:%M')}")
print(f"{'='*65}")
print(f"Total suspicious events: {len(alerts)}")
for a in sorted(alerts, key=lambda x: x["bytes"], reverse=True)[:20]:
    print(f"  [{', '.join(a['flags'])}] {a['domain']} — {a['method']} {a['bytes']}B")