Biometric Data Privacy Hygiene

#!/bin/bash
# Prerequisites: apt install libimage-exiftool-perl ffmpeg (or brew install exiftool ffmpeg on macOS)
#
# Comprehensive metadata stripping for biometric privacy hygiene
# Removes EXIF, IPTC, XMP, GPS, and device-specific metadata

echo "=== Photo Metadata Hygiene ==="

# Strip ALL metadata from a single image
exiftool -all= image.jpg

# Batch strip entire folder (preserve file dates)
# ⚠ -overwrite_original permanently destroys metadata — back up originals first if needed
exiftool -all= -overwrite_original -P ./photos/*.jpg

# Preview what would be removed (dry run)
exiftool -all -s ./photos/sample.jpg

# --- Targeted removal (keep copyright but strip GPS/device) ---
exiftool -gps:all= -Make= -Model= -SerialNumber= \
  -Software= -LensModel= -overwrite_original ./photos/*.jpg

# --- Video metadata stripping ---
# FFmpeg strips most metadata during re-encoding
ffmpeg -i input.mp4 -map_metadata -1 -c:v copy -c:a copy output_clean.mp4

# --- PDF metadata stripping ---
exiftool -all= -overwrite_original document.pdf

# --- Verify metadata is clean ---
echo ""
echo "=== Verification ==="
exiftool -G1 -s output_clean.mp4 | head -20
echo ""
echo "GPS tags remaining:"
exiftool -gps:all ./photos/*.jpg | grep -i "gps"

# Expected output:
# [Photos] Stripping EXIF from photos/*.jpg...
#     3 image files updated
#
# [Video] Remuxing video without metadata...
#     input.mp4 → clean_input.mp4 (metadata stripped)
#
# [Verification]
#     ExifTool Version Number  : 12.76
#     File Name                : photo1.jpg
#     GPS Position             : (none)  ✓
#     Camera Make              : (none)  ✓
#     Date/Time Original       : (none)  ✓

#!/usr/bin/env python3
# No external dependencies — uses Python stdlib only
# ⚠ Data broker URLs and response times are approximate and may change — verify current opt-out URLs before submitting
"""Data broker exposure audit — identify where your biometric data may be stored.
Automates the discovery and opt-out request process for major data brokers."""
import json
import csv
from datetime import datetime
from dataclasses import dataclass, asdict
from typing import List, Optional

@dataclass
class DataBroker:
    name: str
    category: str           # 'people-search', 'background-check', 'advertising', 'biometric'
    data_types: List[str]   # e.g., ['photo', 'address', 'phone', 'social']
    opt_out_url: str
    opt_out_method: str     # 'web-form', 'email', 'mail', 'phone'
    verification_needed: str # 'email', 'id', 'none'
    response_time_days: int
    gdpr_applies: bool
    ccpa_applies: bool
    notes: str = ""

# Major data brokers with biometric/photo exposure risk
BROKERS = [
    DataBroker("Spokeo", "people-search", ["photo", "address", "phone", "social", "email"],
               "https://www.spokeo.com/optout", "web-form", "email", 3, False, True),
    DataBroker("BeenVerified", "people-search", ["photo", "address", "phone", "criminal"],
               "https://www.beenverified.com/app/optout/search", "web-form", "email", 7, False, True),
    DataBroker("Whitepages", "people-search", ["photo", "address", "phone"],
               "https://www.whitepages.com/suppression-requests", "web-form", "phone", 3, False, True),
    DataBroker("PimEyes", "biometric", ["face-embedding", "photo", "source-url"],
               "https://pimeyes.com/en/opt-out", "web-form", "email", 14, True, True,
               "Face recognition search engine — HIGH PRIORITY for biometric privacy"),
    DataBroker("Clearview AI", "biometric", ["face-embedding", "photo", "social-source"],
               "https://clearview.ai/privacy/requests", "web-form", "email", 45, True, True,
               "Law enforcement face recognition — opt-out may not apply to LE use"),
    DataBroker("ThatsThem", "people-search", ["address", "phone", "email", "ip"],
               "https://thatsthem.com/optout", "web-form", "none", 5, False, True),
    DataBroker("Radaris", "people-search", ["photo", "address", "phone", "property"],
               "https://radaris.com/page/how-to-remove", "web-form", "email", 14, False, True),
    DataBroker("FastPeopleSearch", "people-search", ["address", "phone", "email"],
               "https://www.fastpeoplesearch.com/removal", "web-form", "none", 2, False, True),
]

def generate_audit_report(brokers: List[DataBroker], output_file: str = "broker_audit.csv"):
    """Generate opt-out tracking spreadsheet."""
    with open(output_file, "w", newline="") as f:
        writer = csv.writer(f)
        writer.writerow([
            "Broker", "Category", "Data Types", "Opt-Out URL", "Method",
            "Verification", "Expected Response (days)", "GDPR", "CCPA",
            "Request Sent", "Response Received", "Status", "Notes"
        ])
        
        for b in brokers:
            writer.writerow([
                b.name, b.category, "; ".join(b.data_types), b.opt_out_url,
                b.opt_out_method, b.verification_needed, b.response_time_days,
                "Yes" if b.gdpr_applies else "No",
                "Yes" if b.ccpa_applies else "No",
                "", "", "Pending", b.notes
            ])
    
    print(f"Audit report generated: {output_file}")
    print(f"Total brokers: {len(brokers)}")
    
    # Priority ranking
    biometric = [b for b in brokers if "face-embedding" in b.data_types or "photo" in b.data_types]
    print(f"Biometric exposure risk: {len(biometric)} brokers")
    print("")
    print("Priority opt-out order (biometric data first):")
    for i, b in enumerate(sorted(biometric, key=lambda x: "biometric" in x.category, reverse=True), 1):
        print(f"  {i}. {b.name} ({b.category}) — {b.opt_out_method}")

generate_audit_report(BROKERS)

# Expected output:
# Generated: broker_optout_tracker.csv
#
# Priority | Broker        | Category      | URL                          | Status
# ─────────────────────────────────────────────────────────────────────────────────
# 1        | Spokeo        | people_search | spokeo.com/optout            | pending
# 2        | BeenVerified  | people_search | beenverified.com/opt-out     | pending
# 3        | Acxiom        | data_broker   | isapps.acxiom.com/optout     | pending
# ...
# Total: 24 brokers | Estimated completion: 45-90 days

#!/usr/bin/env python3
# No external dependencies — uses Python stdlib only
"""Identity compartmentalization audit — detect cross-account linkability.
Maps shared identifiers that could link separate personas together."""
import hashlib
from collections import defaultdict

def audit_compartmentalization(accounts):
    """Analyze identity isolation between accounts.
    
    accounts: list of dicts with keys:
        persona, platform, email, phone, ip_used, browser_fingerprint, 
        payment_method, display_name, photo_hash
    """
    # Group all identifiers by persona
    personas = defaultdict(lambda: defaultdict(set))
    
    for acct in accounts:
        persona = acct["persona"]
        # photo_hash: SHA-256 hash of profile photo file — generate with: hashlib.sha256(open(photo,'rb').read()).hexdigest()[:12]
        for key in ["email", "phone", "ip_used", "browser_fingerprint", 
                     "payment_method", "display_name", "photo_hash"]:
            if acct.get(key):
                personas[persona][key].add(acct[key])
    
    # Cross-persona linking analysis
    print("=== Compartmentalization Audit ===")
    print("")
    
    all_personas = list(personas.keys())
    link_risks = []
    
    for i, p1 in enumerate(all_personas):
        for p2 in all_personas[i+1:]:
            shared = {}
            for identifier_type in personas[p1]:
                overlap = personas[p1][identifier_type] & personas[p2][identifier_type]
                if overlap:
                    shared[identifier_type] = overlap
            
            if shared:
                risk_level = "HIGH" if len(shared) >= 2 else "MEDIUM"
                link_risks.append({
                    "persona_a": p1, "persona_b": p2,
                    "shared_identifiers": shared,
                    "risk": risk_level
                })
    
    if link_risks:
        print(f"⚠ Found {len(link_risks)} cross-persona linkage risks:")
        for risk in link_risks:
            print(f"  [{risk['risk']}] {risk['persona_a']} ↔ {risk['persona_b']}")
            for id_type, values in risk["shared_identifiers"].items():
                masked = [v[:4] + "***" for v in values]
                print(f"    Shared {id_type}: {', '.join(masked)}")
    else:
        print("✓ No cross-persona linkage detected")
    
    # Per-persona summary
    print("")
    print("=== Per-Persona Identifier Inventory ===")
    for persona, identifiers in personas.items():
        total = sum(len(v) for v in identifiers.values())
        print(f"  {persona}: {total} identifiers across {len(identifiers)} types")

# Example audit
accounts = [
    {"persona": "work", "platform": "LinkedIn", "email": "john@company.com", 
     "phone": "+1555123", "ip_used": "100.1.2.3", "photo_hash": "abc123"},
    {"persona": "work", "platform": "GitHub", "email": "john@company.com",
     "ip_used": "100.1.2.3", "photo_hash": "abc123"},
    {"persona": "personal", "platform": "Instagram", "email": "j_doe@gmail.com",
     "phone": "+1555123",  # SAME PHONE — linkage risk!
     "ip_used": "100.1.2.3",  # SAME IP — linkage risk!
     "photo_hash": "def456"},
]

audit_compartmentalization(accounts)

# Expected output:
# === Persona Compartment Audit ===
#
# ⚠ CRITICAL: Shared email found across 2 personas
#   • work, social → user@gmail.com
#
# ⚠ WARNING: Matching photo hash across personas
#   • work, dating → a1b2c3d4e5f6
#
# ⚠ WARNING: Phone number reused
#   • work, social → +1-555-0100
#
# Linkage risk: HIGH — 3 cross-persona identifiers detected