This doesn't look especially safe. In addition to the fact that the crypto is delivered by the server, and so every browser/server transaction is an opportunity for the server to surreptitiously backdoor the crypto operations, the underlying crypto here appears to be CBC+HMAC where the payloads are decrypted before the HMAC is checked.
Firstly, thanks for inspecting the source and offering a suggestion for improvement. The latter point should be a relatively easy fix.
As for your first point, there’s little we can do to prevent TLS or AWS tampering. But we can make it easier to choose e2e encryption in the first place. So we focused on reducing barriers to entry (no signup required, simple URL-based rooms) as well as providing these benefits over alternatives:
- Open source code
- Ephermeral message history (not persisted in a DB)
- Opt-out anonymity
We think these features make Darkwire a good solution for many users seeking secure, private online communication. Having said that, no solution is perfect and we hope to see contributions from the open source community to make it even better.
Given that the server distributes the keys, and there is no way for the user to compare/verify known-good keys of other users, it's possible for the server to MITM everything, actively, meaning modification instead of just eavesdropping.
>- Open source code
Open source doesn't save you either, because a user cannot inspect what is actually running on the server.
And even the client side, while the code is technically there (downloaded to the browser in minified form), it's not feasible for a user no matter what level of skills to verify the distributed code is actually what's open sourced in your github repository and that the code stays like that whenever you refresh or reopen the site.
So basically, you need to trust the server (operator) and your webcrypto scheme does not provide any additional security whatsoever given that the server (operator) can MITM it trivially. The only thing preventing third parties other than the server (operator) in a MITM position to do harm is the TLS, but not your webcrypto scheme.
>- Ephermeral message history (not persisted in a DB)
Even with best intensions, what happens when you receive some National Security Letter, court order or similar, which forces you into silently MITM the key exchanges, thereby enabling you to MITM the messages and then log every message and pass that to some state actor?
To say something positive, I like the minimal UI and the way you handle file uploads. Even tho your service cannot fulfill the strong user-to-user security it somewhat claims, it's still a nice chat tool.
However, since your service is easy to use (copy-paste just one URL and type away), I'd guess soon enough you'll run into the same troubles as similar services where some bad users (e.g. from the chans) will use your service to e.g. distribute child pornography, revenge porn and other very illegal content.
Thanks for the feedback. If we assume the server is compromised then it's true that a MITM attack is trivial. However it seems to me this would be the case for any web-based e2e chat application, all of which must use a server by definition.
Regardless, it's easy enough to spin up your own instance of Darkwire (`docker compose`) and operate the server yourself if so inclined.
Hey, it is painfully obvious that you have a high school level understanding of the crypto at play here. That's really OK, crypto is hard.
You'll want to look into how real cryptographically secure open source comms apps do end to end encryption. Properly implemented, the server can be fully hostile and never recover messages.
Then you'll need to go remove every claim of e2e or cryptographic security from darkchat. Thank you for your time.
My point was that if the server providing client code is compromised, it can serve malicious code to said client. This isn’t a cryptographic claim, just a point about how all web-based applications work.
Also while I appreciate the feedback, this comment struck me as more hostile than helpful, so I’d suggest having a look at HN comment guidelines for future reference.
Long story short: writing HMAC code that works correctly is fairly easy, writing Encryption code that works correctly much harder, which means there are way more chances a bug will appear. You don't want to send wild unchecked data to that piece of code, because an attacker might be able to exploit the total lack of checks and extract some information about the plaintext, so you really want to verify it comes from a trusted party before munching it.
What you describe is MtE (Mac Then Encrypt), which is generally not recommended because you have to decrypt ciphertext before you can check the MAC. This can lead to various timing or padding oracle vulnerabilities.
The recommended, more safer and sturdier method is EtM (Encrypt then MAC), in which you encrypt the plaintext then MAC the ciphertext.
This way you can verify the cipher before you operate on it.
EtM is [according to wikipedia] "[...] the only method which can reach the highest definition of security in AE [...]".
Being able to supply arbitrary data into your system and have said system blindly apply a secret key to it is IMO not the safe mode. The safe mode is verifying that the data is from a source we can at least moderately trust to now spew out garbage and has the secret key anyway.
I implemented a protocol that uses MtE on a CFB mode cipher. And that's why I asked that question.
Now, it seems like I may did it horribly wrong, I'm going to fix it right now.
Thank you!
EDIT:
After I read an answer[0] on StackExchange, I realized I maybe did it right (Because of the CFB mode is different than CBC). Looks like I need to learn a whole lot more before start that fixing.
Next time I'll just use GCM and save all these troubles.
In the spirit of learning about these things, you might want to try these: https://cryptopals.com/
In particular, problem number 17 from set 3 is an attack against CBC mode AES used with Mac-the-Encrypt. Maybe a fun exercise would be to first implement that attack against CBC mode, and then try to see if you can make it work against your CFB-HMAC protocol.
The best I can recommend is to take NaCl and then just using secret box. That picks the good defaults for you. Anything after that needs some research as even seemingly harmless naive implementations can ruin the entire cryptography of a program.
I don't know what "not the worst choice" means in this context. MtE compositions are generally vulnerable. Are there more flagrant ways to be vulnerable? Sure, I guess?
I am really interested in knowing how you reached your conclusions based on seeing the js file. I am quite new to analysing the source files of a given website using the firefox web console. I tried copying it to a text editor but, man, it was a mess to read. Any pointers as to how I might proceed ?
However, in this case, I can analyze the source now because it is available to me. What should I do when the source isn't available ? How do I proceed when all I have is the 'minified output'?
You can try to prettify/beautify (something like http://jsbeautifier.org/), but the output can still be hard to read. If you're looking to learn how to build a particular feature that's closed-source, I'd recommend trying to build something similar yourself using tutorials or an open-source analog.
Without looking deeper into the app ... The WebCrypto standard makes a lot of assumptions in terms of underlying security and is dangerous (if not negligent) without proper security headers (XSS, CSRF, CSP), in place[0]. Since this site positions itself as a security relevant app, mistakes like these are incredibly worrying.
Neat app, but curious why you aren't using WebRTC p2p?
I've always thought something like this using WebRTC + a (simple) way to independently check that no data is being sent elsewhere would be really cool.
Thanks! We actually need the server intermediary to handle storing and locking rooms, and adding and removing members. Also IIRC WebRTC data channel doesn't have great cross browser support.
- WebRTC is great in theory, terrible in practice (doesn't work very well), and still needs a bootstrapping server. So it is better to have a reliable websocket based server as the default/fallback, and WebRTC progressively enhanced.
- Signal and Whatsapp, as others have pointed out, are far from being comfortable as being private (they already know too much, phone number, etc.), and don't have the convenience of a browser based app.
- Yes, this app should be using the Web Crypto API (it is what we switched over to, for our P2P cryptographic user accounts: https://github.com/amark/gun/wiki/auth ) to reduce dependency on the server. You can use the `integrity` attribute to help out with this, but ultimately unless somebody installs it as an Electron app or something, browser based crypto has its limitations.
> - Signal and Whatsapp, as others have pointed out, are far from being comfortable as being private (they already know too much, phone number, etc.), and don't have the convenience of a browser based app.
Whatsapp does have a browser based app. web.whatsapp.com
This is because the private key to en/decrypt messages never leaves your phone. So the 'browser based' version still routes messages to your phone first (to presumably re-encrypt them with a different key that you got from scanning that QR code).
You can argue semantics if that counts as browser based or not, but I think that's actually a pretty sane way to to things, security wise.
The name makes it sound dodgy: "dark" reminds me of "dark web" and all the negative connotations. Not great for trustworthiness which is especially important for crypto.
How does it differ from wire.com in terms of privacy? Besides, wouldn't you have to transfer the chatroom name over another chat service since it changes all the time and users have no identity of their own?
Darkwire is more for ephemeral chat rooms, where the name doesn't change as long as the room is occupied. We also don't require sign up, so you can use it instantly and anonymously.
You could give the option to the owner of the room to set a password for anyone wanting to join. And maybe you could use the password to encrypt further communication.
I understand the benefits from instant web chat, but I think it's not too difficult to access the signal protocol through Whatsapp or Signal's desktop app.
Those implementations both require phone numbers and registration through a mobile app (hence lack anonymity) and are useless for many use cases.
I think Matrix (Riot) would be a better choice. It's e2e encrypted like Signal but doesn't have any nonsense reliance on phones or phone numbers. It also has extra features like true multiple device support, a federated protocol and an open server implementation.
