Hacker News Reader

Comment by shrimp_emoji | Hacker News Reader

This is not offline, but offline seems silly given today's technological options:

> Both of you generate public/private key pairs

> Both of you spam your public keys to the world (including each other)

>> You might meet offline once, to make sure the public keys you gave each other were indeed yours (and not a MITM's sending theirs to both of you while suppressing your actual messages to each other)

>>> This is probably where your actual question's parameters would come into play, but it only needs to happen once

> You encrypt (using a good cipher) some message into a binary blob with your friend's public key and send the encrypted message

>> Only your friend can decrypt this, using their private key which only they have

> Your friend encrypts their reply with your public key and sends the encrypted message to you

>> Only you can decrypt this, using your private key which only you have

E2EE.

View on HN

> This is not offline, but offline seems silly given today's technological options:

Agreed. There aren't many places where you'll be technically limited in today's world.

However, let's say your're in prison and wants to communicate privately to the outside world. Or, you're in a place where the government is actively/extensively monitoring all technology mediums (to the extent that they could even get your private keys). What are your best options?

The prison example is clearly the most difficult. Under oppressed governments, you may still be able to use digital help to create offline communication parcels. When you're in prison, you have your brain, a pen, paper and if you're lucky, a secure destruction mechanism that doesn't immediately raise suspicion.

You also have lots of time ... that’s definitely an advantage.

>you have your brain, a pen, paper and if you're lucky, a secure destruction mechanism that doesn't immediately raise suspicion.

It would be absurdly difficult, but you could manually implement Diffie-Hellman key exchange to generate a secret with the other party without actually ever communicating it.

Just find a large prime, p (at least 600 digits to be safe).

Then pick an integer g between 1 and p and a random integer x.

Then compute g^x mod p.

Send p, g, and (g^x mod p) to your friend (you can even include these instructions for them -- it won't affect whether this works).

They take your p, g, and (g^x mod p) and choose a random integer y

They compute g^y mod p and send that back to you.

You take their (g^y mod p) and compute (g^y mod p)^x mod p. This equals g^xy mod p.

They take your (g^x mod p) and compute (g^x mod p)^y mod p. This also equals g^xy mod p.

You now share a secure secret key, (g^xy mod p), that you can use for encryption in a cipher. This is secure if you both chose a random x and y, which you never shared, and which any attacker would need in order to derive (g^xy mod p) from the four pieces of information they have: p, g, (g^x mod p), and (g^y mod p). The discrete log problem makes it computationally-infeasible to calculate x or y from those values.

Buuuuut then would come the even harder part: encrypting and decrypting messages (being a human cipher suite) with that secret key. :3 It's just rote math at that point, but... let's hope you have a lot of time and paper.

Reply on news.ycombinator.com