What Each Bitcoin Tracing Tool Can Learn About a Wallet, From Free Explorers to Chainalysis
The customer data that leaked from Keystone and Foundation this week shows who bought a hardware wallet, not what it holds. Connecting a name to coins is the job of a different set of tools, from free block explorers to Chainalysis software, each with its own method and its own documented weaknesses.
On October 24, Keystone said a third party had taken its customers' names, contact information, shipping addresses and order details [1]. Two days later Foundation, the maker of the Passport, blamed compromised API keys for phishing emails sent from its official domains, some of them printed with the recipient's home address [2]. Neither company has listed bitcoin addresses among the exposed data. When Ledger went through a similar leak in 2020, its chief executive wrote that there was "no way to make any correlation" between the leaked data and the funds on customers' wallets [3].
That holds for a customer list on its own. The picture changes once the list is combined with other sources. On September 25, the U.S. Court of Appeals for the D.C. Circuit upheld the conviction of Roman Sterlingov, charged with running the Bitcoin Fog mixer, and in doing so laid out how Chainalysis Reactor, the tool used by the government's expert witnesses, groups addresses. One of its three heuristics relies on "information obtained from data leaks, court documents, and data partnerships" [4]. Several layers of tooling sit between a buyer's street address and their coins, and each sees something different.
Block explorers show everything and name no one
What an explorer does not say is who owns an address. Its "Goggles" filters highlight transaction types such as coinjoins and consolidations, but the project notes that these heuristic categories rest on "subjective criteria" [7]. Interpreting a history remains the reader's work.
WalletExplorer and the co-spend rule
Clustering sites move from single addresses to whole wallets, building on a weakness Satoshi Nakamoto flagged in the white paper. Transactions that draw on several prior outputs "necessarily reveal that their inputs were owned by the same owner," and if the owner of one key is revealed, linking could expose other transactions [8].
WalletExplorer applies that rule directly. Addresses co-spent in one transaction are merged into a single wallet, and the merge chains on to any address that later appears alongside them [9]. Its creator, Aleš Janda, gathered names by signing up for services, transacting with them and watching which wallet his coins joined [9]. The site is candid about its gaps. The name database has not been updated "except some very rare cases" since 2016, because Janda went on to work for Chainalysis, where he is paid to find those names and cannot publish them. New transactions take one to two days to appear, only confirmed ones are shown, and an address that was never co-spent stays unnamed, which the site says is more common for addresses holding larger amounts [9]. On the morning of October 26, the page reported data up to block 973,126, mined the previous evening [9].
Crude as it sounds, the approach goes a long way. In 2013 a team led by researchers at the University of California, San Diego made 344 transactions with exchanges, wallet services, gambling sites and other businesses, hand-tagging 1,070 addresses. Combined with the co-spend rule and a second method for spotting change, those tags let them name 2,197 clusters covering more than 1.8 million addresses [10]. The authors cautioned that change detection depends on how wallets happen to behave and is less robust when those habits shift [10].
Chain-analysis firms add what the chain doesn't hold
Chainalysis and its competitors run the same logic at commercial scale. As the Sterlingov ruling describes it, Reactor layers three heuristics. Co-spending comes first. A proprietary behavioral algorithm follows, drawing on change analysis and the "digital fingerprint" each transaction leaves. The third, intelligence-based clustering, brings in outside information [4]. On October 30, Chainalysis itself disclosed an intrusion that exposed the names, emails and roles of its customers' users [17].
The court let the evidence stand even though both experts acknowledged that Reactor has not been peer reviewed and that they did not know its error rate [4]. The defense argued that a manual check pointed to a 20% error rate. The appeals panel relied instead on the trial court's finding that the tool is deliberately conservative and errs toward attributing too little, and noted that Reactor had tied four of the five addresses used in undercover transactions to Bitcoin Fog [4]. Widespread use by law enforcement and industry counted as general acceptance. For someone wondering what is known about their own wallet, the takeaway is that such attributions are probabilistic and cannot be inspected from the outside.
The wallet's own server sees the addresses
One observer needs no leak at all. Light wallets such as Electrum ask a server for the history of every address they hold. The project's documentation concedes that the client "sacrifices some privacy" this way, since the server can reasonably guess the addresses belong to one entity and also sees the client's IP address unless a proxy, VPN or Tor is used [11]. Sparrow's best-practices guide makes the same point in plainer terms. On a public server the balance is visible to whoever runs it, and the IP address discloses a rough location, which is why the guide suggests moving to one's own node, or a trusted person's, as holdings grow [12]. This week Electrum 4.8.3 also closed a hole that let malicious servers divert payments, covered in our report on the fee-bump attack.
Privacy wallets prevent new links but can't undo old ones
Sparrow and other wallets with coin control let users pick exactly which UTXOs go into a payment, so coins from different sources need not be merged [13]. Sparrow's guide also shows how to build a fake two-person coinjoin, a transaction with two equal outputs that leaves an outside observer unable to tell which one is the payment [14]. Those habits protect future spending. They leave the existing record untouched, and once one address in a wallet is tied to a person, the white paper's warning could extend to everything spent alongside it [8].
| Type | Example | What it can see | Where it falls short | Who learns about the lookup |
|---|---|---|---|---|
| Block explorer | mempool.space [5][6][7] | Any address or transaction, with type filters | Names no owners; its heuristic categories are subjective | The instance operator, unless self-hosted |
| Clustering site | WalletExplorer [9] | Co-spent addresses grouped into wallets | Names largely frozen since 2016; one-to-two-day lag | The site operator |
| Chain-analysis firm | Chainalysis Reactor [4] | Clusters from co-spending, behavior and outside data | No peer review or known error rate | The firm and its clients, never the person traced |
| Wallet server | Electrum, Sparrow on a public server [11][12] | Every address and the balance, plus the IP | Sees only what the wallet asks for | Whoever runs the server |
| Privacy wallet | Sparrow [13][14] | The user's own coins and where they came from | Blocks new links, can't erase old ones | No one, on one's own node |
What turns a customer list into a trail
Knowing where someone lives says nothing about how much bitcoin they hold. The risk begins when an on-chain address acquires a name, because a service knows it, an analyst tags it or a server saw it requested from a particular IP. Co-spend and change detection can do the rest. Both companies have since put numbers on their leaks. Keystone counts 25,964 buyer records holding personal data [15], and Foundation about 5,600 mailing addresses and 5,200 phone numbers it treats as potentially exposed [16]. In both cases the data was supposed to be gone, whether through Keystone's failed erasure tool or the old order copies Foundation's redaction never reached [15][16]. Both incidents are covered in our analysis of the Keystone breach and our report on the fake Passport recall. More of our coverage is in privacy and self-custody.
Sources
Source type in parentheses. Web pages accessed on October 26, 2026, except sources 15 and 16, accessed on October 27, 2026, and source 17, accessed on October 31, 2026.
- Keystone (@KeystoneWallet), We owe you an apology, October 24, 2026. Customer data exposed and data the company says was not. (Primary, official X account)
- Foundation (@FoundationHQ), statement on the fraudulent emails, October 26, 2026, and Zach Herbert (@zherbert), thread on the incident, same day. Compromised API keys, personal data in some emails, a detailed report promised. (Primary, company and CEO)
- Pascal Gauthier, Message by Ledger's CEO, update on the July data breach, Ledger, December 21, 2020. No way to correlate leaked data with funds. (Primary, official statement)
- U.S. Court of Appeals for the District of Columbia Circuit, United States v. Sterlingov, No. 24-3161, decided September 25, 2026. Reactor's heuristics, lack of peer review and known error rate, reliability markers the court accepted. (Primary, court opinion)
- The Mempool Open Source Project, GitHub repository. Open-source code and recommendation to self-host. (Primary, project documentation)
- mempool.space, Privacy Policy, updated July 31, 2024. IP and request logging, deleted after ten days. (Primary, legal document)
- mempool.space, FAQ, "How do Mempool Goggles work?". Transaction-type filters and heuristics based on subjective criteria. (Primary, official documentation)
- Satoshi Nakamoto, Bitcoin: A Peer-to-Peer Electronic Cash System, 2008, section 10. Multi-input linking and the risk once a key's owner is revealed. (Primary, founding document)
- WalletExplorer, Info about WalletExplorer.com. Co-spend clustering, how names were found, name database not updated since 2016, one-to-two-day lag, latest block processed. (Primary, official site)
- Sarah Meiklejohn et al., A Fistful of Bitcoins: Characterizing Payments Among Men with No Names, Internet Measurement Conference, 2013. 344 transactions, 1,070 tagged addresses, 2,197 named clusters. (Primary, academic study)
- Electrum, Frequently Asked Questions. Privacy given up to the server and IP visibility. (Primary, official documentation)
- Sparrow Wallet, Best Practices. Balance visible to the server, IP and location, own-node recommendation. (Primary, official documentation)
- Sparrow Wallet, Frequently Asked Questions. Coin control from the UTXOs tab. (Primary, official documentation)
- Sparrow Wallet, Spending Privately. UTXO model, chain-analysis heuristics and fake two-person coinjoin. (Primary, official documentation)
- Keystone (@KeystoneWallet), Statement Regarding the Security Incident on Our Official Website and Customer Guidance, X article, October 26, 2026. 25,964 buyer records with personal data and the failed erasure tool. (Primary, official X account)
- Foundation (@FoundationHQ), Investigation Update 2, October 26, 2026, and updates on its forum the same day. About 5,600 mailing addresses and 5,200 phone numbers potentially exposed; order copies left in legacy tables. (Primary, official X account and forum)
- Chainalysis, Chainalysis Security Incident Information, October 30, 2026. (Primary, company)
More in Privacy and self-custody
Last updated: October 31, 2026