Linkage attack¶

import pandas as pd
from faker import Faker  # For generating synthetic data
import hashlib

# Initialize Faker for realistic fake data
fake = Faker()

# --- Generate Synthetic Datasets ---
def generate_dataset(num_entries=100):
    """Create two datasets with overlapping attributes for linkage demonstration."""
    # Dataset 1: "Anonymous" forum posts with metadata
    forum_data = []
    for _ in range(num_entries):
        user_id = hashlib.sha256(fake.uuid4().encode()).hexdigest()[:8]  # Pseudonymized ID
        forum_data.append({
            "user_hash": user_id,
            "post_time": fake.date_time_this_month(),
            "topic": fake.random_element(elements=("politics", "gardening", "tech")),
            "location": fake.city(),
            "age_range": fake.random_element(elements=("20-30", "30-40", "40-50"))
        })

    # Dataset 2: "Public records" with similar attributes
    public_data = []
    for _ in range(num_entries):
        public_data.append({
            "full_name": fake.name(),
            "residence": fake.city(),
            "age_range": fake.random_element(elements=("20-30", "30-40", "40-50")),
            "voter_id": fake.uuid4()[:6],
            "occupation": fake.job()
        })

    return pd.DataFrame(forum_data), pd.DataFrame(public_data)

# --- Linkage Attack Simulation ---
def linkage_attack(forum_df, public_df):
    """Demonstrate how overlapping attributes can reveal identities."""
    # Step 1: Find common cities and age ranges
    common_locations = set(forum_df["location"]).intersection(set(public_df["residence"]))
    potential_matches = []

    # Step 2: Cross-reference attributes
    for loc in common_locations:
        forum_users = forum_df[forum_df["location"] == loc]
        public_records = public_df[public_df["residence"] == loc]

        for _, forum_row in forum_users.iterrows():
            matches = public_records[
                (public_records["age_range"] == forum_row["age_range"])
            ]
            if not matches.empty:
                potential_matches.append({
                    "forum_hash": forum_row["user_hash"],
                    "public_name": matches.iloc[0]["full_name"],
                    "location": loc,
                    "age_range": forum_row["age_range"],
                    "confidence": "High" if len(matches) == 1 else "Medium"
                })

    return pd.DataFrame(potential_matches)

# --- Run Simulation ---
if __name__ == "__main__":
    # Generate and link datasets
    forum_df, public_df = generate_dataset()
    matches = linkage_attack(forum_df, public_df)

    print("\n--- Linkage Attack Results ---")
    print(f"Matched {len(matches)} pseudonymous users to public records:")
    print(matches.head(3))

    # Export for analysis (optional)
    forum_df.to_csv("forum_data.csv", index=False)
    public_df.to_csv("public_data.csv", index=False)
    matches.to_csv("linkage_results.csv", index=False)