Everything on a public chain is public forever. That sentence sounds like a slogan until you look up a wallet and find someone's salary, their entire trading history, and everyone who has ever paid them. Privacy tooling exists to close that gap, and almost all of it is narrower than its marketing suggests.
What privacy on a public ledger actually means
Three different things get called privacy, and conflating them is why the debate is so confused.
Address privacy means an observer cannot link your activity together — the payments you receive do not all land at a labelled address. Transaction privacy means an observer cannot see the amount or the counterparty. Network privacy means the observer cannot connect your IP address to your transactions at all.
Most on-chain tooling addresses the first, some the second, and almost none the third. A wallet that hides your balance while broadcasting from your home IP to a centralised RPC has not given you privacy; it has given you the appearance of it.
Mixers: breaking the deposit-to-withdrawal link
A mixer breaks the link between funds that went in and funds that came out. CoinJoin coordinates several users into one transaction so no observer can map inputs to outputs. Tornado Cash instead used a Merkle-tree deposit plus a zero-knowledge withdrawal proof, so a withdrawal demonstrated pool membership without revealing which deposit it came from.
The regulatory response defines the risk today. Tornado Cash was sanctioned in August 2022 — the first designation of a smart contract rather than an entity. Blender.io and Sinbad followed, both tied to the Lazarus Group. Enforcement went beyond sanctions: the Bitcoin Fog and Helix operators were convicted of money laundering, and the Samourai Wallet founders were charged. Exchanges blacklist funds that have touched a sanctioned mixer, which is the practical consequence for a user regardless of intent.
There is a legitimate case — personal security for anyone whose address is public, business confidentiality for a company paying contributors, political dissent where receiving funds is itself dangerous — and the problem is that these sit in the same pool as illicit flow. A mixer cannot separate them.
Stealth addresses: one address per payment
Stealth addresses attack a narrower and far more common problem: address reuse. ERC-5564 derives a fresh one-time address for each payment from the recipient's published stealth meta-address. The sender computes the address and publishes an announcement event; the recipient scans announcements with a viewing key and spends with a spending key. Bitcoin's Silent Payments are the equivalent construction without a registry, and Umbra is the main Ethereum implementation.
What it hides is address linkage, which defeats casual block-explorer snooping and any analytics that clusters a recipient's inflows. What it does not hide is the sender, the amount, or timing correlation — a payroll run in stealth addresses is still visibly a set of equal-sized payments arriving together.
Two implementation costs are real. Scanning announcements is the UX problem: a wallet must check every new announcement against its viewing key, fine on a chain with good indexing and painful without it. And address poisoning is live — a sender publishing a look-alike meta-address can harvest misdirected payments, so wallets must verify what they are sending to.
Privacy pools: the middle path that matters
The most interesting development is the attempt to be private and auditable. Privacy Pools and Railgun-style designs let a user prove their funds did not originate from a designated list without revealing which withdrawal is theirs. Proof-of-innocence schemes like this are the first credible answer to the regulator's objection, because they allow a compliance claim to be made cryptographically rather than by deanonymising everyone in the pool.
Zero-knowledge proofs are the machinery: zk-SNARKs and zk-STARKs prove a statement about hidden data. SNARKs are small and cheap to verify but need a trusted setup; STARKs need none and are quantum-resistant but produce larger proofs. Which you want depends on whether the verification cost or the trust assumption is the binding constraint.
The practical stack
Privacy is a stack, not a feature. A private RPC so your provider is not correlating you. No address reuse. Disciplined funding paths so a fresh address is not trivially linked to a labelled one. And only then, if the threat model warrants it, a protocol-level tool.
For a creator the honest use case is mundane rather than radical: not hiding income, but not publishing a wallet balance and a client list to anyone who looks up an ENS name. Stealth addresses and a fresh address per client achieve that with none of the regulatory exposure a mixer carries. ERC-5564 is an explicit standard, not an evasion tool, and the difference is not cosmetic.
References
Next, read On-Chain Identity 101. More in Security & Privacy.