Know exactly how linkable
your accounts are.
Specter runs state-of-the-art LLM-based deanonymization against your own public profiles and returns a precise map of every signal that ties them together.
No credit card required · audit yourself in under 5 minutes
Built on the research that made deanonymization practical.
In 2026, Lermen et al. showed that LLMs can link pseudonymous accounts at scale from public posts alone. Specter runs the exact same pipeline — ESRC: Extract, Search, Reason, Calibrate — so you learn what an adversary would learn, before they do.
Cross-platform ingestion
Fetch posts, bios, and metadata across Reddit, Hacker News, GitHub, and Stack Overflow through public APIs.
LLM feature extraction
A filter pass removes noise; a capable model pulls identity-relevant micro-data — the exact attack surface attackers exploit.
Two-stage verification
Fast selection followed by deep reasoning. We report the same confidence an attacker would arrive at — with the evidence.
Defensive by design
Specter only scans accounts you control. Raw text is discarded after analysis. No cross-user matching, ever.
Three steps. No surprises.
Provide your handles
Add up to five public accounts you want audited. No passwords, no OAuth.
Pipeline runs
Ingest, extract, embed, reason, and calibrate run sequentially. Typical time: two to five minutes.
Read the report
Exposure score, linked pairs with evidence, leaked attributes, and a prioritized remediation plan.
Pay for what you audit.
Simple, predictable tiers. Cancel anytime.
- ✓Batch audits
- ✓Continuous monitoring
- ✓SSO + API + SLA
- ✓Dedicated support
Audit yourself before anyone else does.
A five-minute Specter run shows you exactly which signals link your accounts — and exactly which edits drop your exposure score fastest.
Start your audit