Support RSA signature verification

[axic]

I propose to support RSA signature verification through a precompiled contract with appropriate fees.

With the current EVM, verification for very low key lengths is trivially implemented using the native 256 bit arithmetic, although it is useless due to inadequate security. Supporting bigger key lengths is prohibitive cost wise. (Requires implementing a bignum library for mul and mod on top of EVM. Alternatively it can be offloaded to an oracle, where trust can be a problem and is also suboptimal.)

Motivation: many PKI schemes, including those employed by government ID cards, rely on RSA. Supporting an easy way to verify a signature would mean to support authentication/authorization using those schemes in a smart contract.

High level method: rsaverify(msg, N, e, S, paddingScheme), where

• msg is the message hash,
• N is the public key modulus,
• e is the public key exponent
• and S is the signature.

Returns a boolean.

Regarding padding schemes I would definitely include an option none, where no padding would be applied and the caller would be expected to handle that.

Challenges: Where this gets complex is defining which key lengths and padding schemes to support and how to define the API. Size of N could be used to determine key length. Possibly the fee would depend on the key length.

[axic]

Here is a sample implementation: https://github.com/axic/ethereum-rsa

And a blog post explaining the situation (a few use cases) a bit more.

[pipermerriam]

This would be pretty excellent. I've been mulling over how to build an identity bridge between ethereum and keybase and this sort of signature verification would help enable that.

[vbuterin]

I would recommend just adding a whole suite of bigint arithmetic precompiles then; more generalized that way.

[coder5876]

This is very cool! We're going to be linking RSA keys to uPort identities probably at first using an external oracle that verifies the signature, but this would be very nice!

[murraci]

This would be very sweet.

[Gustav-Simonsson]

+1 for bigint arithmetic precompiles from which RSA can be easily be built without significant overhead compared to direct RSA op codes

[Renaud-V]

I add my vote to this proposal. The support of RSA signature verification will be a bridge to all applications build on electronic Identity Cards. Estonia, Germany and Belgium already use electronic ID card.
This support will ease/automate lots of administrative processes in countries having eID.

[whitj00]

Hi there,

I was wondering what the status is on this EIP is and how I would be able to help. This is something that would be very useful for a project I'm working on

[aakilfernandes]

+1. Had a cool idea for a microtipping service for developers using ssh keys.

I would love to do npm install some-project --thankyou. Npm or whatever package manager would look at the repo, see whos been making commits, and then send som eth to a smart contract. The smart contract would require an rsa signature based on the ssh keys used in the project in order to claim the funds.

[tristanhoy]

+1 for generic bigint capability! This would allow the implementation of just about any crypto on the Ethereum network.

The particular use-case I'm looking at is where an Ethereum contract is an "oblivious witness" to a particular type of transaction - it can verify the authenticity/integrity of the transaction, but doesn't know the contents of the transaction. Without bigint support, I can't cryptographically verify the transactions within the contract, which means I can't guarantee the state of the contract, which means the state becomes "optimistic" - ALL transactions are accepted and it's up to the client connecting to the contract to figure out which ones are actually valid. Nightmare.

Absolutely critical bigint methods:

• add/sub/mul/div
• pow
• mod
• modExp
• eq/gt/gte/lt/lte

And the following, if you're supporting decimals:

• floor/ceil

The above primitives can be used to trivally construct all of the following algorithms supporting numerous cryptographic applications:

• GCD/EGCD
• LCM
• modInv
• isPrime
• nthRoot

[Alexander-Herranz]

Great idea!