Browsers, operating system and peers can become compromised.
Ultimately users have to be responsible with how and who they connect to. The app otherwise is only as secure and as restricted as any other website. As a web app there are nice features like being able to inspect network activity and code.
A typical mainstream browser can be considered to have been sufficiently reviewed. But you aren’t limited like you would be if the offering was from an app store.
I’m genuinely curious why neither have a webapp offering. You can avoid the official app stores by providing things like APK, but as a webapp you can avoid the installation step, which seems it might be useful for people who would use briar or simplex.
The way this app works, makes it so “anonymous” chat isn’t possible. With IP’s being shared it isn’t a good idea.
I also have quite an ugly UI compared to those other solutions. This will improve over time. But im sure it’s a barrier to attracting users compared to other chat apps.
It depends on how and what you use it for.
WebRTC shares details like ip addresses and if you don’t want those exposed, you shouldn’t use this app. This app is aiming to use true P2P as a step towards true privacy and security.
The app is far from finished.
I’m not sure about what you’re asking. Maybe it’s about getting started? https://youtu.be/KKpu2rXvqfM?si=iWyTv9PzSpcAnsmr
What does “authentication” mean if there’s no server? - the app uses browser based cryptography functions as described here: https://github.com/positive-intentions/cryptography/blob/staging/src/stories/components/Cryptography.tsx … basically asymmetric and symmetric keys are generated between peers on the initial connection and stored on device (indexedDB). maybe this helps: https://positive-intentions.com/docs/research/authentication/
How do browsers behind NAT connect to each other? - the app is using peerjs and so it also uses the peerjs-server as a connection broker. im investigating things like exchanging webrtc connection data offline with things like qr-codes.
How does it verify that the other chat partner is who they say they are? - the asymmetric keys exchanged after the initial connections. i cant drive home a point more clearly. the first connection should be secure, the peerID is cryptographically random, but i have to defer the responsibility of exchanging this ID to a peer they trust. https://positive-intentions.com/docs/basics/getting-started#security-reminders
Why use this and not Simplex? - this app is a work in progress and not ready to replace anything.
this is a side project and im unable set anything aside for having security professionals take a look. its important to note, i am not a cryptography expert… i just know enough to create the app. i try to make this clear in all of my posts that it is for testing purposes only because it could be irresponsible to advertise this this fully working. while the security attempt is genuine. to fix various issues throught the app, i expect there will be breaking changes.
further more about security assessment; while the app is open source, i found that its too complicated for a security assessment without a budget. this is going to be addressed in a ground up implementation of the p2p framework. https://github.com/positive-intentions/p2p … this will eventually replace what is being used in the app and will make it easier to inspect how it works.
thanks for the feedback!
the app is a work in progress and full of bugs and issues. as a side project i can only set aside so much time to do things. the app as you see it is something im approaching with a “release often” approach. this will highlight thing like you are doing for me to prioritize. the profile loading from file is something i havent taken a look at in a while so thanks for that! i will take a look when i can make time for it :)
the cryptography in the app can be a whole separate discussion. but to be brief, the crypto signature input is something the app uses to generate a cryptographically random value… basically whatever you draw (e.g. “a smiley face”) gets converted to a base64 string… then this is passed through a sha-256 hashing function. the idea is that even if you try, you could never draw the “exact” same smiley that would be pixel perfect and result in the same hash. the app also appends this to the cryptographically random value generated from the browsers built-in functions. so that crypto-signature is actually entirely redundent but could address concerns about the device/browser cryptography functions being compromized.
for the QR, i tried to make it a resolution that would be reasonable to view from another phone… to see the data that is being squashed into the QR codes for offline, you can take a look at the details around here. https://github.com/positive-intentions/chat/issues/6 … there is an offline demo described there where instead of QR code it shows the data as plain text (this will include things like IP so you shouldnt be pasting this publicly anywhere… these details generated only exist on your device in memory)
Matrix is a good implementation and I like how it works. Mine is a work in progress and far from finished. It isn’t ready to replace any app or service. The key detail about my app is that it’s browser based. This has its own limitations with what a webapp can do. I think it makes for a different approach to decentralized chat.
How can I describe the chain of trust in a self hosted system? I’m sure auditing will help inspire confidence but this isn’t something I can do for the app. Open sourcing is the next best thing to open it up to public review.