HOW WE
DECIDE.
Every TrustSniffer verdict is built the same way: collect evidence, weigh it with documented rules, produce a score, and land on a clear band, with the reasons shown. No mystery numbers. Here's the method behind both the website and the wallet engines.
Jump to the scoring engineSignals in, one honest score out.
Illustrative example, sample signals and weights, not a real report.
Capture, rules, score, verdict.
Capture
Eight modules collect evidence, page, behavior, archive, DNS, certificate/WHOIS, threat intel, and more. See coverage.
Rules
Deterministic, documented rules weigh each signal. It is not a single black-box AI call, the logic is repeatable.
Score
Signals combine into a 0–100 trust score plus a confidence level, with the contributing reasons retained.
Verdict
The score lands on a band and a plain-English verdict, published with the evidence behind it.
A blocked, challenged, or content-thin capture is graded NEEDS REVIEW, never CRITICAL, a temporarily unavailable but legitimate site is never mislabeled a scam. Scoring today is rule-based; machine-learning calibration is held until its dataset is ready.
Roots, taint, exposure.
How a wallet is scored
We start from known bad roots, sanctioned and scam-linked addresses and issuer-freeze records, then trace how funds flow to and from an address (taint propagation and exposure). Verdicts land on CLEAR · CAUTION · WARN · DANGER · BLOCK, with the connecting evidence (paths, amounts, distance) attached.
Flag vs. confirmed
A flag is an early risk signal, high exposure or a freeze candidate, often ahead of any freeze. A confirmed freeze/unfreeze is an on-chain issuer event with transaction and block evidence. Monitoring alerts on both. A freeze reflects status at a point in time and an issuer can reverse it.
How the Risk Index is counted.
The Risk Index is a set of automated counts over records TrustSniffer has already published. It adds no new score and no new verdict: it counts existing wallet pages and website reports by the record or band each one already carries.
Wallet figures
These are counts of records, not wallet risk scores. Each wallet is counted once, by the reason it has a public page.
- Wallets assessed
- Every supported Ethereum or TRON address with a public TrustSniffer wallet page.
- Sanctioned / frozen (ground truth)
- Addresses with an active record from an outside source: a sanctions listing, a stablecoin issuer freeze, or a reported scam address. When an issuer unfreezes an address or a list removes it, it leaves this count.
- Wallets flagged as risky
- The ground-truth addresses above, plus addresses linked to an adverse entity record.
- Analysed on request
- Addresses someone checked. They count toward the total only, never toward a risk figure.
Website figures
Each website is counted once, using its latest published report. Private reports are excluded.
- Websites with reports
- Every website with a published TrustSniffer report.
- Trust distribution
- The band from that report's 0 to 100 trust score, where 100 is the most trusted: the same bands shown in the website section above, including NEEDS REVIEW for blocked or incomplete captures.
- Websites with risk reports
- Reports in the Low trust or Critical risk bands.
Sources, review and citation
- Who supplies the ground truth
- Government sanctions lists (US Treasury OFAC and Israel's NBCTF), freeze events recorded on-chain by the USDT and USDC issuer contracts, and scam addresses reported in public threat feeds such as Scam Sniffer. TrustSniffer collects these records. It does not create them.
- Outside review
- No outside party audits these figures. You can check them one record at a time: every counted wallet page and website report is public and shows its own evidence. Corrections follow our corrections policy.
- As-of date
- The Risk Index shows the UTC date of its latest count, which is recomputed automatically several times a day. If one part cannot be refreshed, the previous figures stay in place rather than dropping to zero.
- How to cite
- Compiled automatically by TrustSniffer, with no individual author. Cite as “TrustSniffer Risk Index, trustsniffer.com/stats” with the as-of date shown on the page. A plain-text copy is at /stats.md.
The rules behind the rules.
Evidence-first
Every verdict cites the reasons behind it. If we can't show why, we don't assert it.
Deterministic
The same evidence yields the same logic. AI assists wording, not the hard verdict.
Fail-safe
Missing or blocked evidence lowers confidence and routes to review, it never invents a harsh verdict.
Fresh & re-checked
Results can change as new on-chain events, source data, and re-scans arrive.
A TrustSniffer result is a probabilistic, informational assessment from the evidence we could collect at the time, not proof of fraud, not a guarantee of safety, and not legal, financial, or compliance advice. Read it alongside our disclaimer, and see exactly what we support on the coverage page.
See the method in action.
Run a check and read the reasons behind the verdict for yourself.