4

On top of the data exchanged through the Ethereum network, my app stores some extra user data in a classic DB (E.g. thumbnails, display preferences, newsletter management etc.).

Doing that has led me to use a double account system: an Ethereum account—I get it from web3.eth.accounts[0]—and a classic login/PW account. While you can make all sort of links between the two, it still requires to type 2 different PW and manage two accounts.

So I wonder if it is possible to login into the app using just the Ethereum PW (in Mist, MetaMask etc.):

  • Is safe to login into a classic account with just the web3.eth.accounts[0] and no PW?
  • Is it possible to ask for the user permission (like the "accept" prompt with Metamask) but without mining a transaction (too long for a login)?
  • Is there another way to enable single-click login?

The use case is a Meteor app used with MetaMask, but I guess it's a more general question.

|improve this question|||||
5

The knowledge of a public key (such as Ethereum addresses) should never be used as method of authentication.

Public keys are meant to be shared, for instance to receive a payment in a cryptocurrency. If a client can just provide that address to autenticate, then everybody who knows the public address could impersonate the client.

The following JavaScript code should be enough to impersonate an account in this case:

web3 = {eth : { accounts : ["<anExistingPublicAddressInHex>"]}}

after that instruction, the web3.eth.accounts[0] instruction will return the address "<anExistingPublicAddressInHex>" without the knowledge of the private key.

The use of the private key is required to identify the owner of a known public key, for instance by requesting the signature of a string of random data.

|improve this answer|||||
  • 1
    Thanks for posting this. This an important point. I've edited my answer to reflect this. – Samuel Hawksby-Robinson Jun 1 '17 at 11:33
  • 1
    I'm only proposing to use web3.eth.accounts[0] as a means to establish the account we are dealing with. It would be akin to supplying a username. My edited answer details the need for using private key signing as the proof, ie the password. – Samuel Hawksby-Robinson Jun 1 '17 at 12:00
  • 1
    Thanks @SamuelHawksby-Robinson , your answer is great now! I have deleted my comment saying that it was not and that it introduced a security issue. – atfornes Jun 1 '17 at 14:04
2

I'm building a similar system. Some data stored on the blockchain, some stored in a traditional DB, and my personal approach may interest you.

Is safe to login into a classic account with just the web3.eth.accounts[0] and no PW?

The question that you are really asking is "Can you trick web3 into believing you control the private key of an address when you really don't?".

From my understanding and my own attempts at trying to trick web3 the answer is no. This would mean that using the existence of web3.eth.accounts[0] would be proof that the user has the private key of the address.

Yes it is possible to trick web3 into believing you own a public key you do not have the private key for. See atfornes' answer https://ethereum.stackexchange.com/a/16970/2673

So Is safe to login into a classic account with just the web3.eth.accounts[0] and no PW? , I would say yes, but you will need to prove that the user controls the associated private key via a method of signing. See How can I sign a piece of data with the private key of an Ethereum address?

Is it possible to ask for the user permission (like the "accept" prompt with Metamask) but without mining a transaction (too long for a login)?

If you want the user to confirm that they would like to be validated as a user with their web3.eth.accounts[0], you can simply store this data in your traditional DB. I don't see any advantage storing this in a contract. But you will need the user to prove they own the public address by signing some data with the related private key.

Is there another way to enable single-click login?

One of the beauties of using a platform like Ethereum is that you can build a ZERO click login.

If we establish that the user's web3.eth.accounts[0] is proof that the user controls the private key of that account's address, then you will always know that the user is valid.

But you will need some kind of client side script to collect proof that the user has the private key for the public address representing their account.

|improve this answer|||||
  • 1
    Thanks, I wait for a more sure repsonse to accept. My though was that a plugin could maybe emulate the web3 API (like fake MetaMask). Since you only read from the API in this case, you have no way to check they have the private key. The contract would enforce the blockchain interaction (eg. MetaMask. pop-up). But again I'm not sure at all. Agree on the zero click login, that's the point idd! – Etienne Feb 22 '17 at 20:54
  • @Etienne That's fair enough. I had another idea for how I could try and trick web3 into accepting an account I don't have the private keys for, it would be an important discovery as there are currently zero click login systems. Ethlance.com for example uses this approach. – Samuel Hawksby-Robinson Feb 22 '17 at 21:11
  • as I explained in another answer: ethereum.stackexchange.com/a/16970/2214, public keys should never be used for authentication. This answer is wrong and can be dangerous to follow it. – atfornes May 31 '17 at 16:33
2

Recently came across a project in alpha: https://fortmatic.com

Docs seems pretty good, was able to get things working pretty fast. It triggers a modal in an iframe for user auth (similar to Stripe or Plaid) - imo this feels like a better ux as the end users never have to leave the main dapp experience

|improve this answer|||||
  • 1
    Thanks! I didn't try it, but this part seems to do exactly what I wanted then developers.fortmatic.com/docs/user-signing – Etienne Dec 17 '18 at 10:18
  • Hahah I'm a bit late to the response, but still glad to hear that it was what you wanted before – Evan H. Dec 17 '18 at 19:51
1

Had you seen this Auth0 module? https://github.com/auth0/ethereum-authentication-server

|improve this answer|||||
  • Whilst this may theoretically answer the question, it would be preferable to include the essential parts of the answer here, and provide the link for reference. – Afr May 31 '17 at 19:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.