SUTRADHAR links anonymous personas to a single identity across eight independent signals - writing style, behavioural reasoning, timing, identifiers, OPSEC, crypto, infrastructure, and evasion - and shows the evidence behind every link.
A single threat actor runs many aliases across many forums. Watching them is easy; proving that three usernames are one person is not; it takes an analyst weeks of manual cross-referencing.
People carry unconscious tells that survive a username change: how they punctuate, which filler words they reach for, the hours they're awake. SUTRADHAR reads independent signals and fuses them into one knowledge graph.
Every link comes with its reasons. An investigator can see exactly which habits matched before acting on a lead, and the score is a lead, never a verdict.
Each requirement in the problem statement is built and demoable in the live prototype. Nothing is left on the roadmap.
The 2013 Silk Road takedown is the most documented dark-web de-anonymization in history. Every manual signal the FBI used over two years, SUTRADHAR applies automatically in minutes.
What took the FBI two years across multiple agencies, SUTRADHAR runs simultaneously and fuses into one confidence score with a court-ready case file. Sources: U.S. v. Ulbricht (2015), FBI criminal complaint, DEA case documentation.
SUTRADHAR performs genuine HTTP crawling: real requests, real HTML parsing, real breadth-first link discovery across 61 pages and 5 marketplaces. The target is a simulated dark-web network we built ourselves, not a live .onion site.
Every analysis exports as a court-ready PDF, a machine-readable JSON for integration, and a CSV for analyst spreadsheets. All include every signal, every score, and a cryptographic audit trail.
End-to-end protection. At rest: AES-256 (Fernet, PBKDF2-HMAC-SHA256 at 390,000 iterations), ciphertext unreadable without the passphrase. In transit: TLS 1.3. Any tampering fails HMAC authentication and is rejected, never silently accepted. Live proof in the prototype.
The working prototype ships all eight live signals, a 3D knowledge graph, live crawler, visual match, offline AI brief, AES-256 encryption proof, and a court-ready case file.
Live at web-production-9880e1.up.railway.app · local: pip install -r requirements.txt then python app.py