Show HN: Diffie-Hellman exchange for the layman

Diffie Hellman Key Exchange is performed over a cyclic group. That's just a large set of elements that wrap back around to each other. So for example the integers mod 7 is {0, 1, 2, 3, 4, 5, 6}. A generator for this group is any element that you can continuously add to generate the whole group, for example: 1.

In the case of DH, g is some such element.

maple3142• 5 years ago

I think it should also explain why this is secure. It may not be obvious that why you can't just compute the logarithm and get the shared key.

boris119• 5 years ago

I'm the author. If you are referring to my page, the Developer notes provide all the info you need, and the source code is trivial. I've placed all the encryption stuff in a barebones open source library that you can read on GitHub. This makes the remaining source code in the page trivial.

I don't think I can explain to a layman without math backround why it is secure. So I have kept it simple.

happytoexplain2• 5 years ago

I have no idea what this means, personally. I did understand the math and its practical implications long ago: I started with the paint-mixing example, which helped a lot, and went from there.

libeclipse5• 5 years ago

Perhaps a simpler way of explaining:

DH relies on the discrete logarithm problem. That means that if I give you g + g + .. + g, where g is added to itself n times, it will be difficult for you to find out what n is.

Both parties share the same prime p, and the same base point g. Then each of them select a random, secret point. Say Alice's secret is a and Bob's is b. Now each of them create their "public key", which is just g + g + g .. a times, and g + ... b times to create A and B, respectively. The (mod p) simply means divide the result by p and take the remainder as the answer.

Now if I am Alice, my secret is a, my public key is A. Bob sends me B = g^b (mod p). I exponentiate this with my own secret key to get shared secret = (B)^a = (g^b)^a = (g^a)^b = g^(ab) (mod p)

(mod can be deferred to the end of the calculation or done at each step, without changing the answer)

So you can see that, since the order of exponentiation doesn't matter, when Bob does the same calculation (multiplying his private b with alice's public key g^a) he gets the same shared secret.

boris119• 5 years ago

Note that I'm using the Elliptic Curve Diffie-Hellman (ECDH) exchaneg, which is an additive group. That's how I can get the shared keys down to a small size.

anon_tor_123456• 5 years ago

A group is a group. Doesn't matter whether you call it additive or multiplicative. So DH doesn't change since either way g^a and a*g are both defined as repeated action of g on g via the group operation.

ECDH has smaller keys because the attacks are (until now) weaker and not because you're using an additive cyclic group.

sorokod• 5 years ago

p needs to be prime because...?

boris119• 5 years ago

I'm using Elliptic Curve variant. There's no "mod". But you got the general idea. Also, all the math is built-in and done natively by the browser.

NiceWayToDoIT2• 5 years ago

Only thing I would improve on this page is to remove all remote dependencies, youtube, font, cdn files ... sw.js consonantly sending some data. Personally I do not trust Google they have way to more data, and even if all scripts are stripped questions is does Chrome collecting data as well?

libeclipse5• 5 years ago

https://borisreitman.com/privacy_bare.html

boris119• 5 years ago

I'm the author. No data is sent after dependencies have loaded. The "init()" function in the page runs right after. So you can use the Network debugger to see that nothing is sent anywhere.

I just don't think an ugly page can be useful for the layman. It has to be a nice and user friendly page, which means it would have a lot of design loaded as well. Use the "_bare" page if you want to customize it for your needs.

boris119• 5 years ago

You are right about YouTube embed. The sw.js is sending stuff, and I don't want that. I'm going to put the YouTube into an overlay, and will remove it from DOM after it's watched. EDIT: it's done.

NiceWayToDoIT2• 5 years ago

So, yeah I was right only not sure why did I got so many negative points, for stating obvious truth ?! :D

ineiti• 5 years ago

Interesting, but why does the "Check" use a handmade hashing algorithm that is just summing up all bytes?

A much safer implementation would be to take the sha256 of the key and only show the 2 rightmost bytes.

boris119• 5 years ago

Good idea. I will change it, should be live in 1 hour. I am going to SHA-256 hash it, then add the bytes of the hash and mod by 10000. It's better than just taking the first two bytes, because I want a 4 digit number.

MauranKilom2• 5 years ago

It bears note that this is still subject to a MITM attack on whatever channel is used for the information exchange. You cannot generate a shared secret without either an existing shared secret or some form of secure authenticated communication channel in the first place. I mean, you can, but you can't be sure who it is that you're sharing the new secret with.

TickleSteve• 5 years ago

(https://security.stackexchange.com/questions/91699/why-cant-...)

TLDR: "Without authentication, impersonation is feasible, ..."

libeclipse5• 5 years ago

tldr: preventing MITM attacks requires setting up an authentication workflow. Without this, you have no guarantees about who you're speaking to.

For example, TLS 1.3 does this by (after performing ephemeral DH key exchange) signing the conversation transcript with the server's long term identity key. After this the client is sure that they are speaking to the correct server, but the server has no authenticity guarantees about the client.

boris119• 5 years ago

That's why you have the Check Digits. There's no way to communicate a long and cryptic password on the phone. But it is easy to compare check digits on the phone.

MauranKilom2• 5 years ago

I'm not convinced that transmitting the first, say, 5 characters of the full password is any less secure than transmitting the sum of all its "digits". In fact, I would expect the latter to be significantly easier to match by brute force by a MITM adversary.

kabylake999• 5 years ago

cant help but be distracted by there being a circular picture of the author staring at me as i read this

boris119• 5 years ago

I'm the author. This is a good suggestion, and I have moved my picture from the side menu, to the top of the scrollable portion.

thunderbong• 5 years ago

This is really great! I liked the video explanation as well as the page which shows how we go about exchanging information using DH exchange. Will be sharing this page with some of my non-technical friends.

Thank you!

boris119• 5 years ago

Thank you, I am the author. Feel free to contact me if you have tips how to make it even simpler.

boris119• 5 years ago

I've simplified the page by creating two modes: "simple" and "advanced". By default "simple" is shown, and it's what you'd need if you are the recipient, rather than an initiator would would be normally more technical.

libeclipse5• 5 years ago

Note: P-256 is not considered a safe curve. Use Curve25519 instead. See: https://safecurves.cr.yp.to/

boris119• 5 years ago

Hi, I am the author. I know, but I wanted to use something built-in in the browser. When the browser implements Curve25519 I will change the code.

doggodaddo78• 5 years ago

IIRC, aren't there a bazillion gotchas and practical problems implementing DH where it's definitely something one shouldn't roll their own?

boris119• 5 years ago

I didn't roll my own. I'm using built-in browser functionality, available under crypto.subtle object of the browser. You can review the code in my WebUtil library on Github @borisreitman. It's just a light wrapper on top of browser's functionality.

boris119• 5 years ago

I've made a new addition: now the operating system of the other party is also sent as part of the exchange. This info helps decide what software to use for further encryption. e.g. DMG or BitLocker? And, I managed to do it without making the string longer. I stuffed it in the unused bits of the compressed coordinate.

boris119• 5 years ago

I've simplified the page by creating two modes: "simple" and "advanced". By default "simple" is shown, and it's what you'd need if you are the recipient, rather than initiator.

boris119• 5 years ago

I've improved the encrypt/decrypt form. Now the textareas expand, and there's a copy to clipboard button in order to copy the output. It makes it much easier to use on the phone.

boris119• 5 years ago

Update: the tool now allows to encrypt (and decrypt) files. It works in mobile browser too.

Show HN: Diffie-Hellman exchange for the layman

Related Stories