How to Detect Sybil Wallets: A Guide for Crypto Teams

Key Takeaways
Sybil detection is the process of finding wallets that one person or team controls, so that a reward or a vote meant for one user is not counted many times.
It is an identity resolution problem. Two signals do most of the work: shared identity (the same device or user ID) and a shared funding address (the same account sent each wallet its first gas).
Most first funding comes from exchanges, bridges, and relayers. A method that does not remove them reports groups of unrelated users as sybil clusters.
In our platform data, more than nine in ten funding clusters agree with identity data when the funding address is itself a wallet of the app. When it is an outside address, agreement is much lower, and that is also where careful farms hide.
No method proves that two wallets have one owner. Treat a cluster as a reason to review, hold rewards before you ban, and give real users a way to appeal.
Sybil detection is the process of finding wallets that one person or team controls, so that something meant for one user is not counted or paid many times. In an onchain app it is an identity resolution problem: you link wallets through signals that they share, then review each linked group. The two signals that do most of the work are shared identity, such as the same device or user ID, and a shared funding address, the account that sent each wallet its first gas.
This guide covers every common method, what each one catches and misses, why the funding signal produces false positives unless you handle exchanges and bridges, and how to review a cluster before you act.
What Is a Sybil in Crypto?
A sybil is one of many wallets that a single actor controls and presents as separate users. A group of them is a sybil cluster, a sybil ring, or a farm. The actor's goal is to collect something that is given once per user, or to count as many users when they are one.
The same activity has different names in different places:
Term | Where it is used | What is collected or inflated |
|---|---|---|
Sybil attack, sybil farming | Airdrops, governance, points programs | Token allocations, votes, points |
Airdrop farming | Token launches | Airdrop allocations |
Referral farming, self-referral | Referral and affiliate programs | Referrer and referee rewards |
Wash trading | Trading competitions | Prizes, volume rewards, leaderboard rank |
Multi-accounting, bonus abuse | Exchanges and gaming platforms | Welcome bonuses and trading rewards |
For background on the term and its history, see what a sybil attack is. This guide is about detection.
Why Is Sybil Detection an Identity Resolution Problem?
Identity resolution answers one question: which records belong to the same person? In most software the records are emails, cookies, and device IDs. In an onchain app the record is a wallet address, and creating one is free, instant, and needs no email or phone number.
That is why checks on a single wallet do not work. Each wallet in a farm is new, holds enough to qualify, performs the required action, and breaks no rule on its own. The evidence exists only at the group level, so detection starts by linking wallets.
Teams ask us this in different words. One asked for a map of the wallets that belong to one cluster, and whether bot wallets count as active users. Another saw dummy wallets, all funded from one source wallet, that each claimed a mint. An agency asked how to tell organic referrals from self-referrals. A team that runs a trading competition wanted to separate new traders from farmers. And in one demo the question was simply: "Have you been able to solve the problem of users having multiple wallets?" These are all the same question.
What Methods Are Used to Detect Sybil Wallets?
There are seven common methods. They differ in what they need, what they catch, and how easily a farm defeats them.
Method | What links the wallets | Strength | Weakness |
|---|---|---|---|
Shared identity | The same device or the same user ID in your app | A direct match on your own data | A farm that separates devices is invisible |
Shared funding address | The account that sent each wallet its first gas | Hard for a farm to avoid at scale | Exchanges and bridges cause false positives |
Shared exit address | The address where the wallets send their rewards | Catches farms at the cash-out | Only visible after the reward is paid |
Timing | Wallets funded or active in one burst | Cheap to compute | Launches and campaigns also cause bursts |
Behavior similarity | The same actions, amounts, and order | Finds scripted farms | Real users also follow guides step by step |
Proof of personhood | A score or credential tied to a human | Independent of your data | Coverage is partial, and a low score is not proof |
Published sybil lists | Wallets named in past airdrop investigations | Free and specific | Frozen in time, and limited to those airdrops |
The first two link wallets. The other five add or remove confidence. A strong case uses a linking method and at least one confirming method.
Shared identity
When several wallets connect from the same browser, or sign in with the same user ID, they are probably controlled by one person. This is a direct match on data from your own app and needs no inference.
It has two limits. Two real people can share one device. And a careful operator gives each wallet its own browser profile or its own device, which makes the farm invisible to this method.
Shared funding address
A new wallet cannot do anything until it holds gas, and someone has to send it. That first transfer is public and permanent. The funding address of a wallet is the account that sent its first native transfer on a chain, and wallets that share one form a funding cluster.
This method survives what defeats the first one. An operator can rotate devices, networks, and wallets. Funding each wallet from a different, unrelated source costs time and money for every wallet on every chain, so most operators fund many wallets from one place.
Shared exit address
Farms have to collect their rewards, and many send them to one wallet. This signal is strong, but it arrives late: you see it after you paid. Use it to recover from a past campaign and to design the next one, not as your first line of detection.
Timing, behavior, proof of personhood, and lists
These four do not link wallets on their own. They change how much you trust a link.
Timing. Ten wallets funded within minutes of each other point to one operator. Ten wallets funded over a year from the same outside address point to a service.
Behavior. The smallest qualifying deposit, the same trade size, the same order of actions, then nothing.
Proof of personhood. Scores such as Human Passport, and credentials from verified exchange accounts, help separate people from scripts.
Published lists. LayerZero, Hop, and Optimism published sybil lists after their airdrops. A wallet on one of them has a history.
Why Does the Funding Signal Produce False Positives?
Most first funding does not come from a person. It comes from infrastructure. When a user withdraws from an exchange to a new wallet, the exchange's hot wallet is the funding address. Bridges, relayers, and gas top-up services behave the same way. One such address funds thousands of unrelated people.
If you group wallets by funding address and stop there, your largest "sybil clusters" are groups of exchange customers. When we built funding clusters in Formo, about seven in ten of the funding addresses that had funded more than one wallet were exchanges, bridges, or other services. Among the addresses that had funded ten or more wallets, nearly all were.
Removing services takes several steps, because each one catches what the others miss:
Public address labels identify known exchange wallets, bridges, and gas services.
Activity identifies services that have no label yet. An account that has sent far more transactions than any person could is a service.
Relayer patterns identify accounts that deploy a separate contract for each user. They look like one person who funded many wallets, but each wallet belongs to someone else.
Reach identifies an address that has funded a very large number of wallets across many apps.
Labels alone are not enough, because labels arrive late. We found a funding address with no public label that had formed a cluster within days. It was an exchange hot wallet only a few weeks old, with tens of thousands of transactions already sent, and its "linked" wallets were spread across many devices and several countries. Its activity gave it away long before any label did.
How Accurate Is Sybil Detection?
Accuracy depends on the kind of link. We compared funding clusters with identity clusters on Formo's platform data. The two signals are independent, one from the chain and one from the app, so agreement between them is a useful test.
Funding address of the cluster | How common | Agreement with identity data |
|---|---|---|
A wallet that also uses the app | Most clusters | More than nine in ten |
An outside address | A minority | About six in ten |
Read the two rows differently.
The first row is the common case: a user who funds their own second wallet. It is reliable, and it is usually harmless. Most people with two wallets are not farming.
The second row is where judgment is needed. Lower agreement does not mean these clusters are wrong. A farm that uses a separate device for each wallet shares no identity by design, so it can only ever appear in this row. This is also where a service that was not yet recognised would appear. That is why a cluster needs a confidence level and a review, not an automatic decision.
One pattern from our data shows the value of the second row. One account funded a batch of wallets within a few minutes. Each wallet then used the app from its own devices over several months, with no device shared between any two wallets. Identity data saw unrelated users. The funding address was the only link.
How Do Farmers Evade Sybil Detection?
Every method can be evaded, at a cost. Knowing the cost tells you what a farm of a given size can afford.
Method | How a farm evades it | What it costs the farm |
|---|---|---|
Shared identity | A separate browser profile or device for each wallet | Low. Antidetect browsers make this routine |
Shared funding address | A separate, unrelated funding source for each wallet | High. Typically one exchange account per wallet, per chain |
Shared exit address | A separate cash-out path for each wallet | Medium. It delays and splits the proceeds |
Timing | Random delays between wallets | Low, but it slows the whole operation |
Behavior | Random amounts and action order | Low to medium |
Proof of personhood | Buying or renting verified identities | High, and it grows with every wallet |
Two conclusions follow. Device-based detection is the cheapest to defeat, so do not rely on it alone. And the funding address is among the most expensive to hide, which is why it catches operators who have already defeated everything else.
How Do You Review a Sybil Cluster?
Review a cluster by checking whether independent signals agree. Work in this order:
Check the funding address. Is it a wallet that uses your app, an outside address, or a service?
Check the funding window. Minutes apart points to one operator. Months apart from an outside address points to a service.
Check for shared identity. Do any two wallets share a device or a user ID? A link that both signals confirm is the strongest.
Compare behavior. Look at first seen dates, countries, devices, and actions. A farm tends to act in one rhythm.
Check the confirming signals. Humanity scores, sybil lists, and wallet history.
Decide the action. Hold the reward and record the reason. Exclude or ban only when several independent signals agree.
What about real users who are caught?
Some will be. Two people in one household share a device. One person sends gas to a friend. A user with a main wallet and a hardware wallet is one person with two wallets, and that is not abuse unless your rules say one reward per person.
Three practices keep this fair:
Hold, do not ban, as the first step. A hold can be undone.
Record the reason for each decision. "Twelve wallets funded by one address within ten minutes" is a reason. "Flagged by the system" is not.
Offer a way to appeal. A real user can usually explain a shared device or a shared funding source quickly.
What Data Does Sybil Detection Need?
Less than many teams expect, and the choice has privacy consequences.
Shared identity needs an identifier in the browser and, if you have logins, your own user ID.
Shared funding needs only public chain data.
Device fingerprinting needs detailed device attributes. It is personal data in many jurisdictions, and it is the signal that a farm defeats most cheaply.
Formo does not fingerprint devices and does not store raw IP addresses. It links wallets through an anonymous ID in the browser, through the user ID that you set, and through the funding address onchain. See device fingerprinting versus identity resolution for the full comparison.
How to Detect Sybil Wallets in Formo
Formo runs both linking methods and shows the reason for every link.
Cluster by identity or by funding address. On the Users page, open the Clusters tab and use Cluster by to switch between the two views.
Confidence level. Each funding cluster shows High confidence or Needs review, with the reasons on hover: whether the funding address is a wallet of your app, how many wallets also share an identity cluster, and within how many days the wallets were funded.
Services are removed. Exchanges, bridges, relayers, and gas services do not form clusters. Formo recognises them from public labels and from their activity.
Every link has a reason. A wallet profile lists the linked wallets with an icon for each reason, and shows the funding address or the service that funded the wallet.
Confirming signals. Wallet labels include Human Passport scores and the LayerZero, Hop, and Optimism sybil lists, and you can filter and segment by them.
Export. Export a cluster as a CSV file with the linked addresses, the reason for each link, the funding address, and the confidence level.
Funding clusters cover the major EVM chains and Solana. See identity resolution in the docs for the method and its limits.
Frequently Asked Questions
What is sybil detection?
Sybil detection is the process of finding wallets or accounts that one person or team controls. It works by linking wallets through shared signals, mainly a shared device or user ID and a shared funding address, and then reviewing each linked group.
How do you detect sybil wallets?
Link wallets first, then confirm. Link them by shared identity (the same device or user ID) and by a shared funding address. Then check timing, behavior, humanity scores, and published sybil lists. Remove exchanges, bridges, and relayers before you trust any funding link.
What is a funding address?
The funding address of a wallet is the account that sent the wallet its first native gas transfer on a chain. Wallets that share a funding address were probably prepared by the same operator, unless the funding address is a service such as an exchange.
Why do exchanges cause false positives in sybil detection?
An exchange pays withdrawals from a small number of hot wallets, so every user who withdraws to a new wallet gets the same funding address. Without a step that removes exchanges, bridges, and relayers, a clustering method groups thousands of unrelated users together.
Is sybil detection the same as identity resolution?
Sybil detection is identity resolution applied to an adversarial case. Identity resolution asks which wallets belong to the same person. Sybil detection asks the same question about people who are trying to look like many.
Does a sybil cluster prove fraud?
No. A cluster shows that wallets are linked, not why. People share devices, and one person can send gas to a friend. Check whether several independent signals agree, hold rewards before you ban, and let users appeal.
Can a farm avoid funding-based detection?
Yes, by funding each wallet from a different and unrelated source, for example a separate exchange account for each wallet. That costs time and money for every wallet, so many operators do not do it. It is one reason to combine funding with identity, timing, and behavior.
Is multi-accounting the same as a sybil attack?
Yes, in method. Multi-accounting is the term used on exchanges and gaming platforms for one person with several accounts. Sybil attack is the term used in crypto for one actor behind many wallets.



