Chain Analytics: What Surveillance Can and Can't See

Bitcoin's whitepaper has a section titled Privacy, and its own conclusion is unglamorous: the ledger is public, addresses are pseudonymous, and the protection breaks once several inputs can be tied to a single owner. That sentence is the foundation of the chain-analytics industry, and it is also its outer boundary. Everything a surveillance vendor reports is either published ledger data or an inference built from published ledger data plus an anchor outside the chain. This piece is about which inferences are hard, which are guesses, and what actually changes the picture: swapping the asset, or swapping the endpoint. Every figure below was measured on 19 October 2026.

The raw material is public and enormous. At 06:00 UTC on 19 October 2026 we pulled the newest Bitcoin block: height 972,110, 3,234 transactions, 1,639,060 bytes, weight 3,993,967 (virtual size 998,491.75 vB). Inside it, 8,755 inputs were spent into 8,235 outputs — 2.71 inputs and 2.55 outputs per transaction on average. 3,110 of the 3,234 transactions were SegWit (96.2%). Fees for the whole block totalled 1,327,367 sat, an average of 1.33 sat/vB, with a median of 1 sat/vB and a range from 0.14 to 60 sat/vB. The coinbase went to AntPool. At the same moment the mempool's recommended rates ran from 2 sat/vB (fastest) down to 1 sat/vB. None of this is private — it is the published record, and every analytics vendor ingests the same file.

Heuristics, not magic. Three techniques do most of the work. First, common-input ownership: inputs signed together in one transaction are almost certainly controlled by one wallet. In a 200-transaction sample from that same block, 21 transactions (10%) carried more than one input, averaging 3.12, and the largest carried 319 inputs — a single transaction that publicly binds 319 addresses into one ownership cluster. Second, change detection: spending a 0.5 BTC output while paying 0.13 BTC usually leaves a change output that is still yours. Third, amount and timing fingerprints. Note the noise level: fee rates inside one block spanned 0.14 to 60 sat/vB, so this transaction resembles that one is a probabilistic claim, not proof. What turns addresses into identity is an anchor, and anchors sit outside the chain. (how Monero fees work)

Where identity enters: the anchors. A pile of addresses is not a person. Identity arrives at the points where the chain touches a regulated system: a KYC exchange withdrawal, where the venue knows the account holder and usually the withdrawal address; a fiat on-ramp; a merchant invoice; or an address published somewhere. Every address downstream of an anchor inherits a label, and if you later deposit back to that venue the label travels the other way, too. This is also why your own habits matter more than any vendor's algorithm: reuse one payout address across five swaps, or reuse one refund address, and you have published the link yourself. Round amounts inside a short time window do nearly the same thing for less effort. (recovering misdirected funds)

Monero removes the amounts and keeps only that something happened. On Monero the same observation yields far less. A public node reported height 3,787,297 with 64,803,456 transactions in history; block 3,787,297 weighed 229,661 bytes with 65 transactions, and block 3,787,290 weighed 62,749 bytes with 30. What you can see is that transactions exist, how heavy they are, and what the network earned in fees — because RingCT has hidden amounts since 2017, and each input is signed against a ring of 16 possible sources, one real and fifteen decoys (a mandatory default since 2022; Monero is working toward FCMP++, which would widen that set toward the whole output pool). An observer cannot tell which input was real or how much moved. What stays exposed is contextual: which node your wallet queries, which address you handed the swap service, and how you behave around both. (how Monero fees work)

The swap page is the seam where two chains get stitched. An instant swap is not a mixer and we do not run one — but structurally it is a place where a Bitcoin-side deposit and a Monero-side payout sit inside the same order record. Anyone watching a service's deposit addresses can pair entry and exit by amount and by timing without knowing who owns either end. That is why the operational habits above matter: if every order uses a fresh payout address you control, a separate refund address, and a non-round amount, the pairing gets weaker. It never reaches zero. A cross-chain swap moves value, not blame. The useful mental model is reduce what is provable, not become invisible. (privacy coins vs mixers)

Six habits that cut what can be proven. 1. Receive payouts into an address you hold the keys for — never an exchange deposit address, which hands the venue your whole swap history. 2. Use a fresh payout address per order; most wallets generate them for free. 3. Avoid round and repeated amounts: 0.04137 leaves a thinner fingerprint than 0.05. 4. Never reuse a refund address between orders — it is the most common self-inflicted link. 5. Separate the network layer: our Tor mirror at 4jyoxqdyx7dh3xfjctihpld7fgliowhimajfszcgypxash5rjqhas3yd.onion hides your IP from the service, and hides nothing about the addresses you type. 6. If chain-level privacy is the actual goal, move to XMR or a shielded ZEC address instead of shuffling between transparent chains. (Is Tor safe for swaps?)

Three claims that do not survive contact with the data. First, analytics firms see everything in real time. They see every transaction; without an anchor they see no identity, and their clustering is heuristic — a cluster is a claim with a confidence level, not a verdict. Second, no-KYC means untraceable. No-KYC removes the account link only; the ledger stays public and services still screen addresses. That decides whether an order is accepted, not whether you are invisible. Third, Tor equals anonymous. Tor protects the network layer and does nothing about address reuse. Two honest limits on our side: we cannot tell you what any specific vendor's label database holds, and we do not support mixing. Pressed for a summary: privacy is a property of a whole path — wallet, network, asset, counterparty — and one strong link does not repair four weak ones. (Is Tor safe for swaps?)

What to do next. Run the test on your own last swap: was the payout address fresh, self-custodied and non-round, and was the refund address used only once? If any answer is no, that is where the provable link came from — not from the chain itself. For scale, measured on 19 October 2026: 0.05 BTC quoted 6.88917858 XMR (137.78 XMR per BTC), 1 XMR quoted 574.632879 USDT-TRC20, spot references were XMR $592.52, BTC $82,681 and ZEC $1,164.69, and our catalog listed 540 assets across 72 networks. The rates above were live at writing — today's numbers live on the swap page. Chain-level privacy starts with the asset you pick and the addresses you type, not with the hop in the middle. (recovering misdirected funds)