Chain Surveillance

Anna pays for a book with bitcoin. She uses a wallet, an application or device for receiving and sending payments. Once the transaction is confirmed, she opens it in a blockchain explorer to look at the public history of confirmed Bitcoin transactions. She sees which earlier outputs her wallet spent and the new amounts it created. An unspent output is called a UTXO: an amount with conditions governing its later use. The transaction’s inputs point to the earlier outputs it spends. Some money went to the merchant; after the fee was deducted, the remainder returned to Anna as change. The public record does not normally label those roles. Nor does it have a mandatory field for a name or IP address. It also does not assign each individual input satoshi, the smallest bitcoin unit, to a particular output. Colored arrows in an analysis tool may therefore include its own value-allocation rule, rather than additional information confirmed by the network. [1] [7] [8]

A wallet can spend several earlier receipts together to make a purchase. An analyst may infer that the same entity controls all the inputs. This clue is called the Common Input Ownership Heuristic; a heuristic is a useful assumption, not a certainty. Guessing which output is change can extend the group further. There are documented counterexamples: CoinJoin lets several participants build a joint transaction, while Payjoin lets a recipient add their own input to a sender’s payment. BIP 78 describes this process directly. A shared transaction alone therefore does not prove common ownership; participants need not share their private keys to create it. A private key is a secret number used to create a signature authorizing bitcoin spending. [2] [5] [7]

The decisive connection may come from outside the blockchain. A merchant knows Anna’s order, an exchange may hold a withdrawal record, and a public fundraiser may publish its receiving address. An analyst can obtain a label connecting an address to a service or person. But its origin and applicable period matter. A connected group of addresses, or cluster, is not a list of people. An exchange may hold funds for many customers, while one person may use several disconnected groups. Labeling an address with an exchange’s name does not establish which customer paid, who had the economic interest in the funds or what was purchased. [2]

Anna leaves more than a transaction behind online. A purchase may be linked to an account or a cookie, a small piece of browser data used to recognize a visitor. In a 2017 study, Steven Goldfeder and coauthors showed how purchase information could be connected to a public payment. The amount, a time window or a payment address can provide clues. Identity did not suddenly become part of the blockchain; two data sources were linked. This historical study supplies no universal success rate for today. It does highlight that stored information may be connected later, so not knowing someone’s name today need not protect them forever. [3]

An IP address identifies a network connection, whereas a Bitcoin address helps specify where a payment can be sent. An observer may record an IP during a connection or transaction relay. This is a different information source from the public payment history. A node is a computer participating in the Bitcoin network. It also forwards other people’s transactions: the first computer an observer hears from need not be the originator. Tor can help hide a user’s direct network connection, but does not rewrite the amounts or links of already published UTXOs. Assessing a conclusion requires knowing what was actually observed on the network and what was inferred from transactions. [6]

Imagine a tool mistakenly assigns the merchant’s address to Anna. If it then merges whole groups using further shared links, the error can eventually affect many people. This mistaken merging of groups is called cluster collapse. Malte Möser and Arvind Narayanan studied ways to limit such mistaken merging, while also noting limitations in their validation data. Analysis quality cannot be reduced to one impressive percentage. False links, missed connections, coverage and the period studied all matter. Without calibration — comparison against known cases — a tool’s score cannot be presented as the probability that a proposed name identifies the right person. [4]

For Anna, this explains the value of new receiving addresses: others cannot connect payments quite so easily just by seeing the same address repeated. Spending those receipts together later can create another clue, however. Payjoin weakens the common-input assumption; it does not erase public transactions or the merchant’s order record. Running a node reduces the need to ask someone else’s server about your addresses, but confirmed payments remain public. Each measure addresses a different connection. Instead of promising complete anonymity, ask what a particular merchant, wallet provider or independent observer can learn about Anna. [1] [5] [6]

Good analysis lets readers trace a conclusion back to its evidence. It identifies the transaction or output, the label’s origin, the assumption used and the date, while allowing alternative explanations. A direct payment differs from a connection through many intervening transactions. Being able to sign need not mean having the economic interest in the funds, as with custody for customers. A risk score or distance in a graph (a map of transactions and their connections) alone therefore proves neither identity nor a person’s conduct. These questions form a review approach derived from the limitations described in the sources. New evidence must allow a conclusion to be corrected, rather than repetition turning the original estimate into supposed certainty. [2] [4]