What is cryptography?
Cryptography is the science of protecting and authenticating data. Cryptography works via encryption, which is the process of scrambling an original message so as to make it unreadable until it’s decrypted by the recipient. Think of the secret codes that children used to send to each other across the classroom. In modern cryptography, encryption systems use math to make “attacks” computationally infeasible.
Encryption is one tool in the cryptographic kit. Another is hashing, which turns any piece of data into a short fingerprint. A fingerprint identifies the data exactly, changes completely if the data changes at all, and cannot be run backward to recover the original.
Cryptography is widely used today in payment systems, online security, and cryptocurrencies. A well-known example is blockchain technology which uses signatures and hashing to create a permanent record of decentralized financial transactions.
What is proof?
Proof is a way to confirm something happened. An event occurred. The sun set. Two people spoke on the phone. Imagine you have hired a baker at a famous bakery to bake her signature carrot cake. How can you get proof that this baker has in fact baked the cake?
You could ask her to send you a detailed video recording of her at work. You could ask your friend to go and check whether she has baked the cake. Or you could drive to the bakery yourself to see if the cake is ready.
These are all reasonable actions. The problem is that none of them are infallible. A video can be edited and falsified. A friend can make mistakes and be confused. And even if you go into the bakery yourself, even if you eat a slice of the freshly baked cake, you cannot be entirely sure it was baked by the baker, using the premium ingredients you paid for, and not outsourced to someone else. These methods–logs, third-party verification, observation–all place trust in systems that are not 100% dependable. When it comes down to it, none of this is real proof.
So what can you do if you need irrefutable proof? Well, you might turn to math. Mathematical proofs rely on logic to guarantee that a statement is true. A proof places trust in math. The benefit of a proof is that unless math breaks down, and unless the fundamental rules of the universe stop working, a proof cannot be broken.
Proof will guarantee that the conditions you set when you made your order–one carrot cake baked by the head baker, this morning, and using only real vanilla–have been carried out.
What is a zero-knowledge proof?
If proof sounds good, zero-knowledge proof is even better. A zero-knowledge proof (ZKP) is a type of mathematical proof that guarantees the existence and select properties of an event without revealing anything else.
To accomplish this selective disclosure, a ZKP relies on the fingerprint tool. The baker takes her secrets, the recipe, the receipts, the name of her butter supplier, and produces a cryptographic fingerprint of them. The fingerprint cannot be reversed to reveal the recipe, and changing even one ingredient would produce an entirely different fingerprint. The proof is a series of mathematical checks performed against that fingerprint, built so the checks can only pass if the hidden facts satisfy your conditions. All that comes back is the yes/no answer: Was the set of conditions satisfied or not?
In this way, the baker can let you know the cake is ready, and she can also guarantee it is the special kind they are known for, without revealing the proprietary recipe she wants to hide from competitors. She can prove that the butter she used is organic, without revealing her exact source.
Is it useful?
Yes! In the course of everyday business, there is certain information you may never want to reveal. Your strategy, position on the market, sensitive client data, or exact operational procedures, for instance. However, financial regulations require a certain level of transparency to prove that you complied. Traditionally, this is handled through detailed logs because proof without encryption relies on observability.
If you want to prove that you sent a payment to a particular person, the only way to do this using observability requires you to show everything you did. This could include every strategy, every move, and every piece of sensitive financial information that was involved in the transaction. Credit card numbers, amounts, locations, etc.
This can inadvertently create vulnerability in the systems you are trying to protect. Compiling all the data in one log, for observation, creates concentrated risk that is vulnerable to breaches. It raises many risks associated with data custody. One breach and suddenly everything is exposed.
This system also reveals the boundaries of authorization and shows who has access to what. Demonstrating that you followed the rules can inadvertently show an attacker how best to target you, who they need to compromise for access.
Zero-knowledge proof means you can prove compliance without exposure. You can show that you followed all the rules without revealing any of your strategy. All the details stay private.
What makes the Inherence zero-knowledge proof unique?
Historically, cryptographic engineers have created zero-knowledge proofs for policy adherence using specialized programming languages, usually Circom or Noir. These are handwritten proofs that take significant time and expertise to generate. And they are susceptible to bugs.
Inherence makes that process automatic and performant. We use an algorithm that is provably correct. The result is that we become the first company to automatically compile policy adherence to a zero-knowledge proof in a scalable way.
Why does that matter?
Handwritten proofs work fine when the rules are static. And few. But if the policy changes, or if the number of policies increase, it quickly becomes impossible to keep up. Every time you want to make a change, you need to get an engineer to come and write a new proof. Policies change all the time; this is inconvenient. Trading strategy responds to the ever-evolving market. Access needs are continually shifting depending on personnel. The Inherence algorithm can turn these policies into proof automatically, without slowing down the workflow.
Another serious downside of handwritten proofs is that they can be buggy, so it’s impossible to guarantee that the proofs are always correct. On a small-scale, this can be monitored. However, in a larger context, agents executing thousands of actions sequentially, themselves spawning hundreds of sub-agents, vastly increases the risk surface in a way that traditional zero-knowledge proof is simply not designed to handle. And, because they take non-deterministic paths to achieve the ultimate objective, it’s impossible to anticipate the source of risk, or the attack vector it will take.
The most important quality in an AI agent is that it does what it’s told. An agent that makes an error by purchasing goods from a non-approved vendor, three hundred times in one minute, is more of a liability than an asset.
Inherence proofs are all provably correct. Like painstakingly handwritten proofs, Inherence proofs are super fast at runtime. Unlike handwritten proofs, each Inherence proof is machine-certified using SMT (satisfiability modulo theories), meaning that each proof will be correct every time. When the stakes are high, and when real money is involved, the possibility of mistakes is not acceptable.
What does this mean for me?
If you are involved in AI, defi, or compliance, this is all good news. Inherence offers the possibility of automating routine financial transactions safely and quickly; think payments, procurement, revenue splits, and countless others. It reduces cost and eliminates friction, increasing the economic pie for all of us. And the transparency and accountability that Inherence provides has the potential to eliminate billions of dollars of waste and fraud.
Have your cake; eat it too!