3 ms·

Ask HN: Is there any way for two machines to talk over an untrusted network?

I realize Kerberos authentication exists, which requires a trusted third party. Is there any way for two machines to communicate securely without a shared secret or a trusted third party?

raimue 9y ago
This depends on your goals. If you only want to communicate, a secure channel with a shared secret can be established with the Diffie-Hellman key exchange. However, this will not authenticate the other party, which means ensuring that you are actually communicating with the expected partner and not with a man in the middle. For this, the partners need to present something only they know. This would usually be solved with asymmetric crypto, like verifying certificates signed by a trusted third-party or using previously seen public keys in a trust-on-first-use scenario.
- apparentiguana 9y ago
  Thanks for the info! I've been reading up on Diffie-Hellman, wasn't sure if it was susceptible to MITM, but glad to know that it is. My goal is to enable a mostly-automated way to have devices on a network discover and authenticate each other. As a thought experiment, I was trying to determine if two devices plugged into a malicious router could communicate securely, but everything I've found so far seems to indicate that it isn't possible without a shared secret. Do you know how web vendors do it? Does Chrome/Firefox/etc have any base keys stored to ensure you aren't working with a censored internet?
  - stephenr 9y ago
    That's what the root Certs are for. Firefox ships with its own, chrome uses the platform ones.
    - apparentiguana 9y ago
      Oh gotcha, thanks! Really appreciate the info, I'll go read up on root certs.
grizzles 9y ago
How web vendors do it: PKI. PKI (root certs) was very controversial at the time it was proposed. It has endured because it's easy for things that are happening to keep happening. It's the idea that because a company with a root cert (eg. Verisign) says you are who you say you are, they will vouch for you. But it's not that secure, and plenty of people have demonstrated over the years it's easy to impersonate companies/websites by swindling Verisign first. Microsoft has been a victim on numerous occasions. Diffie Hellman isn't really susceptible to a MITM. It just means the party at the other end of your secure channel could be a hostile actor. So you need a way to discriminate on who's naughty/nice. Like Santa. For your use case you could distribute a usb stick of blessings to the computers you have deemed virtuous. But a nice computer can turn into a naughty computer, etc. So if the two machines can communicate at least once over a hostile network, then they can ensure future conversations are secure (co-authenticated) and not susceptible to mitm. The problem is, they might not be communicating that first time. I hope this captures the nuance.
- apparentiguana 9y ago
  It does, thank you! I've designed a flow that I think will work based on this and the other comments - I think the understanding that there's not an easy way without a shared secret has been helpful. What I'm going to do is use a shared secret can be used to sign a public key for each machine, since the secret will presumably only be shared with machines I control, as long as I can prevent other security concerns from compromising them.

I realize Kerberos authentication exists, which requires a trusted third party. Is there any way for two machines to communicate securely without a shared secret or a trusted third party?

raimue 9y ago

This depends on your goals. If you only want to communicate, a secure channel with a shared secret can be established with the Diffie-Hellman key exchange. However, this will not authenticate the other party, which means ensuring that you are actually communicating with the expected partner and not with a man in the middle. For this, the partners need to present something only they know. This would usually be solved with asymmetric crypto, like verifying certificates signed by a trusted third-party or using previously seen public keys in a trust-on-first-use scenario.

apparentiguana 9y ago
Thanks for the info! I've been reading up on Diffie-Hellman, wasn't sure if it was susceptible to MITM, but glad to know that it is. My goal is to enable a mostly-automated way to have devices on a network discover and authenticate each other. As a thought experiment, I was trying to determine if two devices plugged into a malicious router could communicate securely, but everything I've found so far seems to indicate that it isn't possible without a shared secret. Do you know how web vendors do it? Does Chrome/Firefox/etc have any base keys stored to ensure you aren't working with a censored internet?
- stephenr 9y ago
  That's what the root Certs are for. Firefox ships with its own, chrome uses the platform ones.
  - apparentiguana 9y ago
    Oh gotcha, thanks! Really appreciate the info, I'll go read up on root certs.

grizzles 9y ago

How web vendors do it: PKI. PKI (root certs) was very controversial at the time it was proposed. It has endured because it's easy for things that are happening to keep happening. It's the idea that because a company with a root cert (eg. Verisign) says you are who you say you are, they will vouch for you. But it's not that secure, and plenty of people have demonstrated over the years it's easy to impersonate companies/websites by swindling Verisign first. Microsoft has been a victim on numerous occasions. Diffie Hellman isn't really susceptible to a MITM. It just means the party at the other end of your secure channel could be a hostile actor. So you need a way to discriminate on who's naughty/nice. Like Santa. For your use case you could distribute a usb stick of blessings to the computers you have deemed virtuous. But a nice computer can turn into a naughty computer, etc. So if the two machines can communicate at least once over a hostile network, then they can ensure future conversations are secure (co-authenticated) and not susceptible to mitm. The problem is, they might not be communicating that first time. I hope this captures the nuance.

apparentiguana 9y ago
It does, thank you! I've designed a flow that I think will work based on this and the other comments - I think the understanding that there's not an easy way without a shared secret has been helpful. What I'm going to do is use a shared secret can be used to sign a public key for each machine, since the secret will presumably only be shared with machines I control, as long as I can prevent other security concerns from compromising them.