Confidential is not the same as anonymous

Privacy is not one property. It is at least three, and each can fail independently.

A public blockchain publishes four things about every payment: who sent it, who received it, how much, and when. Each is a separate disclosure, and hiding one does not hide the others.

Much of the industry says "private" and stops there. That is not precise enough to reason with, and it is not precise enough to buy on.

Three properties, not one

The vocabulary for this predates the industry and is more precise than most of the language used today. The reference terminology, set out by Pfitzmann and Hansen, separates anonymity, unlinkability, undetectability, and several neighbouring ideas that are routinely merged in practice.

Reduced to what matters for payments, there are three.

Confidentiality hides the amount. The transaction is visible, and the parties may be visible, but the value is not readable.

Anonymity hides identity. The amount may be in the clear, but the payer cannot be tied to a real-world party.

Unlinkability hides the pattern. Individual payments may be visible and even attributable, but they cannot be combined into a picture of who deals with whom, how often, and in what rhythm.

A system can provide any one of these without the others. Many provide exactly one.

Confidentiality is the narrowest of the three

Encrypting amounts is the property that has advanced fastest because cryptography handles it relatively cleanly. Homomorphic encryption and zero-knowledge proofs can let a network verify that a balance did not go negative without any node learning what the balance is.

What that buys is specific. A competitor cannot read your revenue, a supplier cannot read your margin, and payroll is not a public spreadsheet.

What it does not buy is anonymity. Encrypting a value says nothing about who moved it. A system that hides amounts and nothing else can still publish an edge: two parties interacted, at a known time, through a known contract.

That is not a flaw in the technique. It is the boundary of what the technique addresses.

Anonymity is a property of the crowd, not of you

Anonymity is not a property of a single user. It is measured against an anonymity set, meaning the population from which you are indistinguishable.

That has an uncomfortable consequence: your privacy depends on other people's behaviour. A pool with ten thousand active participants can offer meaningful cover. The same protocol with eleven offers almost none, and no amount of cryptography inside it changes that.

It also means anonymity can degrade with observation in a way confidentiality does not. An encrypted amount stays encrypted. An anonymity set can shrink every time a participant exits carelessly.

Unlinkability is the one that quietly leaks

The property many systems are weakest on is also the one least often named.

Even when amounts are encrypted and identities are pseudonymous, a payment graph can still reveal a great deal. Timing, frequency, counterparty degree, and the order of interactions are all metadata, and metadata is often enough. Address clustering built on exactly this kind of signal is already a mature commercial industry.

Two payments a month to the same address, on the same day, every month, can tell a story without revealing a single dollar figure.

Why the distinction is not academic

The three properties have very different relationships with regulation, and merging them causes real confusion.

Confidentiality is broadly compatible with regulated use. Amounts can be hidden from the public and disclosed to an auditor or regulator when required, because the data still exists and an authorised party can reveal it. Selective disclosure is a coherent design.

Anonymity is harder to reconcile with that model. If a system genuinely does not know who transacted, it cannot produce that information later for anyone, under any legal instrument. That is a defensible design choice, but it is also the one that has attracted the most enforcement attention.

Treating confidentiality and anonymity as the same thing is how a confidential system inherits the reputation of an anonymous one, and how a buyer ends up believing they purchased a protection they did not.

Where Privara sits

Privara is a confidentiality system. Balances, amounts, and counterparties are encrypted on-chain and remain verifiable. What is encrypted can be disclosed to a regulator or an auditor when the law requires it.

It is not an anonymity system, and describing it as one would be wrong. There is no mixer and no anonymity set. A payment leaves a record; what that record reveals is the part Privara controls.

That is a deliberate position, not an unfinished one. Confidentiality is the property a business needs to protect payroll, margins, and treasury size while still being able to answer an auditor.

The honest limits

Encryption is not the whole of privacy. In any system, a determined observer with timing data and enough off-chain context may still infer something about a business whose figures they cannot read, and no design removes that risk entirely.

The three-way split in this piece is also a simplification of a larger literature that includes undetectability, unobservability, and pseudonymity, none of which reduces neatly to the other three.

Privara is on testnet on Arbitrum today. Privacy properties are easier to state than to preserve under adversarial load, and a testnet is not that load.

The bottom line

When a payment system says it is private, the useful follow-up is: private in which sense?

Ask what a well-resourced observer would still be able to learn. If the answer is "nothing at all," be sceptical. If the answer is precise and bounded, you are talking to someone who has thought carefully about the problem.

Confidential is not anonymous. Both are worth having, and neither is the other.