Understanding Wallet Risk Signals

WalletGenome calculates a 0–100 heuristic score for each selected chain. Four implemented factor groups can add points. The score is capped at 100 and mapped to grades A, B, C, D, or F. A factor appears when its configured threshold is crossed.

• Observed unlimited approvals to unverified contracts can add up to 40 points. When none meet that high-risk rule, more than three observed unlimited approvals to known contracts can add up to 20 points.
• A failed-transaction ratio above 5 percent can add up to 25 points.
• More than two observed approvals older than 180 days can add up to 15 points.
• More than five unidentified contract interactions, combined with an unknown-contract ratio above 30 percent, can add up to 10 points.

When no factor crosses a threshold, the report records that no configured factor crossed its scoring threshold in the returned complete dataset. This statement describes the model and dataset. It does not certify the wallet as safe.

Each chain has its own score, grade, and ordered factor list. The factor list shows the model contribution and a description of the observed condition. When the required history is complete, the portfolio headline uses the maximum selected-chain score and the corresponding worst grade. It does not average a high-risk chain into quieter networks.

Start with the chain behind the headline. Then review every factor on that chain and compare it with the same factor across the remaining selected networks. A score without its chain, records, and coverage state is too compressed for a defensible finding.

Contract addresses, spenders, labels, transaction costs, and activity patterns are chain-specific. The same wallet can use different applications and approval practices on Ethereum, Base, Arbitrum, and Optimism. Record the chain beside every factor before deciding whether several observations describe one repeated behavior.

Approval factors use ERC-20 approval transactions reconstructed from returned history. WalletGenome keeps the newest observed state for each token and spender pair and excludes observed revocations. An old or unlimited approval can raise the score under the configured rules. The report does not query the live allowance, so current on-chain state requires a separate verification step.

Approval exposure answers a related financial question. It estimates the positive token balance covered by the latest observed non-revoked approval, using an available current price and the finite allowance cap when applicable. Unknown balances or prices make exposure unavailable. Approval authority can still affect the score when a monetary exposure estimate is unavailable.

The failed-transaction factor uses the ratio of failed normal transactions. The unknown-contract factor uses contract interactions that the configured classifier cannot identify. Counterparty tables, protocol labels, gas patterns, and transfer flows provide context for those factors. They do not add score points unless they enter one of the implemented scoring rules.

WalletGenome withholds the canonical risk score and grade when wallet history is incomplete. This prevents a partial provider response from producing a definitive headline. Individual returned records can still support a bounded observation, provided the report states that the wider history is incomplete.

• Missing transactions can hide failures, approvals, revocations, or unknown-contract interactions.
• Missing token balances or current prices can make approval exposure unavailable.
• Missing current prices can remove the monetary dimension from the behavioral Sybil heuristic; the remaining dimensions are reweighted.
• An unavailable list check supplies no clear or flagged conclusion for that source.

A low score means the configured factors contributed few or no points within the complete returned history. A low score cannot establish safety. The model does not detect every exploit, compromised key, social-engineering event, off-chain threat, or future action.

A high score means one or more configured factors crossed their thresholds. A high score cannot establish fraud, malicious intent, compromise, or wrongdoing. Legitimate activity can include repeated failures, old approvals, automation, and interactions with contracts that the label set does not recognize.

Use the score to order the review. Use transaction and approval evidence to support the finding.

WalletGenome reports non-behavioral public-list and blacklist matches separately from its local MEDIA-style behavioral heuristic. The blacklist status becomes flagged when a configured non-behavioral source positively matches. The behavioral result uses engagement, protocol and contract diversity, cross-chain breadth, wallet age, and a monetary dimension when pricing is complete.

A source match records that the address appears in a configured dataset. The local heuristic estimates how closely returned behavior fits its scoring pattern. A list match does not prove the current controller’s intent. A behavioral indicator does not prove common ownership, coordinated control, or Sybil participation. A clear list result does not prove absence from every relevant dataset.

1. Confirm completeness

   Record the history, price, approval, and list-check states that support the output.

2. Find the leading chain

   Identify which selected network supplies the maximum score and grade.

3. Trace each factor

   Open the supporting approvals and transactions. Record hashes, timestamps, spenders, labels, ages, and failure ratios.

4. Add context without changing evidence type

   Use counterparties, protocol labels, transfers, gas, identity records, and Sybil states to form follow-up questions.

5. Verify time-sensitive state

   Check current allowances or other live chain state with a trusted chain-specific source when the decision depends on present conditions.

6. Write a limited conclusion

   Describe the observed factor, its chain, its data window, and the uncertainty that remains.

The methodology documentation lists the current model and reporting contracts. Recheck it when a decision depends on an exact rule or threshold.