[Milestone] 4bytes Database

[kuzdogan]

Related: #2171

[manuelwedler]

Some feedback we got:

It's certainly useful for block explorers (like Otterscan) for decoding transactions and events from unverified contracts. Since it's easy to brute force 4 bytes, though, you need to weight the signatures according to usage or have some other system for deciding what's the "true" signature.

I think it could be useful to show the number of verified contracts which use each plain text signature.

[manuelwedler]

Keeping more feedback here:

First major use case: early detection of unverified contract functions.
This is especially important for identifying potentially malicious or scam contracts based on known function signatures (e.g. blacklisting risky patterns before verification).

Additional suggestion: implement a function signature transformer.
Many contracts implement the same core logic with slightly different names or argument orders — e.g., openTrade(), startPosition(), initTrade(), etc. Grouping or clustering similar signatures (via semantic or argument pattern similarity) would greatly increase utility for static analysis, monitoring, and search.

Existing solutions don’t offer a public API—it’s all for their internal use. Most projects have Telegram channels, but nothing accessible via API, and even those channels are often inactive or poorly maintained. The main DB right now out there also not updated. As well we missing labels but i have a list and can send you.

Regarding grouping: since you already have an alliance, you could consider establishing a common standard. I’m actually doing something similar right now, as I’m parsing cross-chain data and adapting to constant changes. I’m happy to help if you’d like.

For example, the term "open trade" is commonly used. Any function with similar logic could be grouped under that label. I’d also recommend adding a column to blacklist certain functions. Fraudulent actors tend to reuse the same function, just with renamed the values, so it’s useful to flag those patterns.
I even tried identifying patterns in encoded functions but had to stop—it became too time-consuming. Still, it could be useful. Even when functions are encoded, there may be recognizable patterns that help detect scam behavior early. Often it's about spotting certain combinations, though that can get complex.

Many tokens launching on EVM aren’t being flagged in time, and users are still aping in. When functions are decoded, detection is much easier. But even for encoded ones, it's possible to catch them if the right logic is in place.

awesome and yes add another column blacklisted true false for function they are truly used to scam. I found alone 700+ but have to go through again as there 100% some duplicates. (That should be careful checked). And yes create a global label field.

[kuzdogan]

Related project: https://github.com/volsa/etherface

[kuzdogan]

Adding feedback:

Something 4bytes doesn't have but would be useful is to include error codes. 4bytes only includes function sigs and events.

For example
error InvalidDecimals(); // 0xd25598a0

[kuzdogan]

Related ETHBerlin Hackathon proposal I wrote a while ago for someone to build this ethb3rlin/find-a-team#32