This is a `playground`

to test code. It runs a full `Node.js`

environment and already has all of `npm`

’s 400,000 packages pre-installed, including `@guildofweavers/genstark`

with all `npm`

packages installed. **Try it out**:

var genstark = require("@guildofweavers/genstark")

- all documents on RunKit are public
`require()`

any package directly from npm- use arrow functions, classes, template strings, and most of ES6
`await`

any promise instead of using callbacks (example)- create your own embedded
**node.js**snippets

This service is provided by RunKit and is not affiliated with npm, Inc or the package authors.

This library is intended to help you quickly and easily generate STARK-based proofs of computation using JavaScript. The goal is to take care of as much boilerplate code as possible, and let you focus on the specifics of your computations.

A STARK is a novel proof-of-computation scheme that allows you to create an efficiently verifiable proof that a computation was executed correctly. The scheme was developed by Eli-Ben Sasson and team at Technion-Israel Institute of Technology. STARKs do not require an initial trusted setup, and rely on very few cryptographic assumptions. See references for more info.

**DO NOT USE THIS LIBRARY IN PRODUCTION.** At this point, this is a research-grade library. It has known and unknown bugs and security flaws.

```
$ npm install @guildofweavers/genstark --save
```

Here is a trivial example of how to use this library. In this example, the computation is just adding 2 to the current value at each step. That is: x_{n+1} = x_{n} + 2.

```
import { instantiateScript } from '@guildofweavers/genstark';
// define a STARK for this computation
const fooStark = instantiateScript(Buffer.from(`
define Foo over prime field (2^32 - 3 * 2^25 + 1) {
secret input startValue: element[1];
// define transition function
transition 1 register {
for each (startValue) {
init { yield startValue; }
for steps [1..63] { yield $r0 + 2; }
}
}
// define transition constraints
enforce 1 constraint {
for all steps { enforce transition($r) = $n; }
}
}`));
// create a proof that if we start computation at 1, we end up at 127 after 64 steps
const assertions = [
{ register: 0, step: 0, value: 1n }, // value at first step is 1
{ register: 0, step: 63, value: 127n } // value at last step is 127
];
const proof = fooStark.prove(assertions, [[1n]]);
// verify that if we start at 1 and run the computation for 64 steps, we get 127
const result = fooStark.verify(assertions, proof);
console.log(result); // true
```

There are a few more sophisticated examples in this repository:

- Demo STARKs - demonstration of how to use various features of this library.
- MiMC STARK - basically the same as Vitalik Buterin's MiMC tutorial.
- Rescue STARKs - various STARKs based on Rescue hash function (e.g. proof of hash preimage, Merkle proof).
- Poseidon STARKs - various STARKs based on Poseidon hash function (e.g. proof of hash preimage, Merkle proof).

When you run the examples, you should get a nice log documenting each step. Here is an example output of running 128-bit MiMC STARK for 2^{13} steps:

```
Starting STARK computation
Set up evaluation context in 146 ms
Generated execution trace in 52 ms
Computed execution trace polynomials P(x) in 7 ms
Low-degree extended P(x) polynomials over evaluation domain in 83 ms
Serialized evaluations of P(x) and S(x) polynomials in 92 ms
Built evaluation merkle tree in 87 ms
Computed composition polynomial C(x) in 574 ms
Combined P(x) and S(x) evaluations with C(x) evaluations in 50 ms
Computed low-degree proof in 231 ms
Computed 48 evaluation spot checks in 2 ms
STARK computed in 1327 ms
--------------------
Proof serialized in 8 ms; size: 94.58 KB
--------------------
Proof parsed in 6 ms
--------------------
Starting STARK verification
Set up evaluation context in 9 ms
Computed positions for evaluation spot checks in 1 ms
Decoded evaluation spot checks in 1 ms
Verified evaluation merkle proof in 4 ms
Verified transition and boundary constraints in 52 ms
Verified low-degree proof in 14 ms
STARK verified in 85 ms
--------------------
STARK security level: 96
```

You can find complete API definitions in genstark.d.ts. Here is a quick overview of the provided functionality:

The simplest way to create a STARK for a computation is to instantiate it from AirScript source like so:

```
const myStark = new instantiateScript(source, options, logger);
```

where:

`source`

is the AirScript source code. If this parameter is a`Buffer`

, it is expected to contain AirScript code in UTF8 format. If the parameter is a string, it is expected to be a path to a file containing AirScript code.`options`

is an optional parameter with additional STARK options.`logger`

is an optional logger. The default logger prints output to the console, but it can be replaced with anything that complies with the Logger interface.

When provided, STARK options parameter should have the following form:

Property | Description |
---|---|

extensionFactor? | Number by which the execution trace is "stretched." Must be a power of 2 at least 2x of the constraint degree, but cannot exceed 32. This property is optional, the default is smallest power of 2 that is greater than 2 * constraint degree. |

exeQueryCount? | Number of queries of the execution trace to include into the proof. This property is optional; the default is 80; the max is 128. |

friQueryCount? | Number of queries of the columns of low degree proof to include into the proof. This property is optional; the default is 40; the max is 64. |

hashAlgorithm? | Hash algorithm to use when building Merkle trees for the proof. Currently, can be one of the following values: `sha256` , `blake2s256` . This property is optional; the default is `sha256` . |

wasm | A flag indicating whether to use WebAssembly optimizations. This proper is optional, the default is `true` . |

**Note:** WASM-optimization is available for certain finite fields and hash functions. If the field or the hash function you are using does not support WASM-optimization, a warning will be printed and its JavaScript equivalents will be used. In general, WASM optimization can speed up STARK proof time by 2x - 5x.

You can also instantiate a STARK from AirAssembly source, or even from an AirSchema object:

```
const myStark = new instantiate(source, component, options, logger);
```

where:

`source`

is either`Buffer`

with AirAssembly source code, a string that is a path to a file with AirAssembly source code, or an`AirSchema`

object.`component`

is a name of the component within AirAssembly source from which the STARK is to be instantiated. If this parameter is omitted the`default`

component will be instantiated.`options`

is an optional parameter with additional STARK options.`logger`

is an optional logger. The default logger prints output to the console, but it can be replaced with anything that complies with the Logger interface.

Once you have a `Stark`

object, you can start generating proofs using `Stark.prove()`

method like so:

```
const proof = myStark.prove(assertions, inputs?, seed?);
```

where:

`assertions`

is an array of Assertion objects (also called boundary constraints). These assertions specify register values at specific steps of a valid computation. At least 1 assertion must be provided.`inputs`

is an array of values initializing all declared inputs. This parameter must be provided only if the STARK requires inputs.`seed`

is an array of seed values for initializing execution trace. This parameter must be provided only if the STARK requires initialization from seed values.

Handling of inputs deserves a bit more explanation.

If you've instantiated a STARK from AirScrip source, you don't need to worry about the `seed`

parameter as AirScript does not support explicit trace initialization. However, STARKs instantiated from AirScript source always require `inputs`

parameter. The shape of this parameter will depend on how you've defined your inputs (see AirScript documentation for more info).

For example, if your input declaration looks like this:

```
public input foo: element[1];
```

your `inputs`

array will need to contain a single element which will be an array of values for input `foo`

. For example:

```
[[1n, 2n]]
```

If you've declared more than one input, like so:

```
public input foo: element[1];
public input bar: element[1];
```

your `inputs`

array will need to contain an array of values for each declared input. For example:

```
[[1n, 2n], [3n, 4n]]
```

Here, the `[1n, 2n]`

values are assigned to input `foo`

and `[3n, 4n]`

values are assigned to input `bar`

.

If you've declared nested inputs like so:

```
public input foo: element[1];
public input bar: element[1][1];
```

your `inputs`

object may look like so:

```
[[1n, 2n], [[3n, 4n], [5n, 6n]]]
```

Here, for each value of `foo`

, we need to provide a list of value for `bar`

.

If you've instantiated a STARK from AirAssembly source, you may need to provide values for both `inputs`

and `seed`

parameters based on what is required by AirAssembly declaration. The `inputs`

parameter requirements are defined via input registers, while `seed`

parameter requirements are defined via trace initializer.

Once you've generated a proof, you can verify it using `Stark.verify()`

method like so:

```
const result = myStark.verify(assertions, proof, publicInputs?);
```

where:

`assertions`

is the same array of Assertion objects that was passed to the`prove()`

method.`proof`

is the proof object that was generated by the`prove()`

method.`publicInputs`

is an array of values for initializing all declared public inputs.

Verifying a proof basically attests to something like this:

If you start with some set of inputs (known to the prover), and run the computation for the specified number of steps, the execution trace generated by the computation will satisfy the specified assertions.

Assertions (or boundary constraints) are objects that specify the exact value of a given mutable register at a given step. An assertion object has the following form:

```
interface Assertion {
register: number; // index of a mutable register
step : number; // step in the execution trace
value : bigint; // value that the register should have at the specified step
}
```

Some very informal benchmarks run on Intel Core i5-7300U @ 2.60GHz (single thread):

STARK | Field Size | Degree | Registers | Steps | Proof Time | Proof Size |
---|---|---|---|---|---|---|

MiMC | 128 bits | 3 | 1 | 2^{13} | 1.3 sec | 95 KB |

MiMC | 128 bits | 3 | 1 | 2^{17} | 23 sec | 147 KB |

MiMC | 256 bits | 3 | 1 | 2^{13} | 11.5 sec | 108 KB |

MiMC | 256 bits | 3 | 1 | 2^{17} | 230 sec | 165 KB |

Merkle Proof (Rescue, d=8) | 128 bits | 5 | 8 | 2^{8} | 300 ms | 60 KB |

Merkle Proof (Rescue, d=16) | 128 bits | 5 | 8 | 2^{9} | 600 ms | 72 KB |

Merkle Proof (Poseidon, d=8) | 128 bits | 8 | 12 | 2^{9} | 900 ms | 74 KB |

Merkle Proof (Poseidon, d=16) | 128 bits | 8 | 12 | 2^{10} | 1.8 sec | 84 KB |

STARKs in the above examples have security parameters set to provide ~96 bits security.

**Note 1:** Rescue and Poseidon hash function instantiations are not really "apples-to-apples" - refer to here and here for exact parameters.

**Note 2:** Currently, STARKs in 128-bit fields are able to take advantage of WebAssembly optimization, and thus, are much faster than STARKs in 256-bit fields.

This library is originally based on Vitalik Buterin's zk-STARK/MiMC tutorial. Other super useful resources:

- STARKs whitepaper: Scalable, transparent, and post-quantum secure computational integrity

Vitalik Buterin's blog series on zk-STARKs:

- STARKs, part 1: Proofs with Polynomials
- STARKs, part 2: Thank Goodness it's FRI-day
- STARKs, part 3: Into the Weeds

StarkWare's STARK Math blog series:

- STARK Math: The Journey Begins
- Arithmetization I
- Arithmetization II
- Low Degree Testing
- A Framework for Efficient STARKs

Other STARK libraries:

MIT © 2019 Guild of Weavers

- package on npmhttps://npmjs.com/package/@guildofweavers/genstark
- licenseMIT

RunKit is a free, in-browser JavaScript dev environment for prototyping **Node.js** code, with every **npm** package installed. Sign up to share your code.

Sign Up for Free