I don’t think it’s possible over a channel that has active interest in breaking your encryption. It’s the recipe for a mitm attack.

It would be possible to use password encrypted keys on the public channel and rotate the keys immediately after establishing a connection.

I try to provide ways to transfer data offline. I don’t have a working example, but it would be possible to exchange keys over a series of QR codes or NFC… But this would require peers to be at the same physical location.

I’m sure there are many approaches for exchanging keys here that balance between convenience and security.

Ultimately it’s worth noting we are talking about encryption on a browser where the operating system could just as easily be taking screenshots.

Im also working on something similar. I was thinking with a browser extension I could “scan” a page for some kind of blob that the extension would have produced. Something like a webcomponent tag (but not that because a good website would have sanitized for that) then inject a webcomponents from the extension to show it decrypted. The idea is pretty difficult to describe without a working example, but I think it would allow for reasonably secure messaging on public channels (because the payload would be encrypted).

As for exchanging keys securely, the app is using crypto-random ID’s to connect peers. On the initial connection, it’ll establish the nessesary keys.

https://positive-intentions.com/docs/research/authentication/

Thanks. Can you tell me more about how a P2P system can get infiltrated. I’m fairly confident in my approach. I hope I’m not being too naive.

Users can always make mistakes/bad-choices and so I’ve tried to create some guidelines to follow.

https://github.com/positive-intentions/chat#security-and-privacy-recommendations

Thanks for your thoughts. I’m sure many others share your opinions.

I’m no expert on that matter, but I assume many people work on open source for various reason. For me, This is something I hope will lead to supporting me.

I find myself recently unemployed (it happens). And so I have to figure out my next steps in life. I’d like to work on this project full-time, yet without funding, I can only continue part-time. I’m motivated to work on the project because I see its still fairly unique, but without any way for it to support me, it’s future on shaky grounds… Especially when faced with issues I can’t ignore like unemployment.

It’s also why I’m reluctant to invite collaborators at this unstable stage because then going close-source becomes immoral.

The app is only using (local-only) browser storage (specifically indexedDB).

So in a P2P interaction, the traditional concept of “the cloud” is just the physical devices connected over a direct webrtc

Thanks for the tip. The future of the encryption is still under consideration. I’m not against the tox protocol… The signal protocol seems better regarded.

thanks for your thoughts. npm is popular for a reason and vanillajs doesnt scale very well. so any deps used could be an issue.

i was also considering if with the webcomponent approach it could be “furture-proof” as it seems to be the rhetoric i hear around. im sure i wont have a great implementation any time soon, but id like to try out a few ideas to see if it holds-up. hopefully to lead to a “secure javascript ui framework” (which itself could be a whole discussion).

but based on all the feedback ive recieved, it seem for the messaging app refactor, i’ll be fine to use react on it. which is great because i already have a working-ish demo.

thanks for your thoughts.

thats not quite what im asking. im wondering if there are nuanced benefits to using webcomponents over something like react. with the key difference being the native support.

i hope with the webcomponent approach it could be “furture-proof” as it seems to be the rhetoric i hear around. im sure i wont have a great implementation any time soon, but id like to try out a few ideas to see if it holds-up. hopefully to lead to a “secure javascript ui framework” (which itself could be a whole discussion). i hope that by having it open source, i can point to an example to discuss and improve it.

it seem for the messaging app refactor, i’ll be fine to use react on it. which is great because i already have a working-ish demo.

https://positive-intentions.com/blog/async-state-management/ https://positive-intentions.com/blog/bottom-up-storage/

the state management with the useStore hook seem tricky to explain and can be a whole separate dicussion. i try to explain a bit in those posts. im still working on it more and while it isnt document anywhere, but im also investigating having it store data encrypted-at-rest.

thanks. thats what id like to aim for and i dont think its far off. the build script there is mainly for the storybook statics (as seen in the link provided for “website”).

couple things i hope to do soon, remove lit as a dependency - i use this right now because its useful for template rendering and lifecycle methods. webcomponents have a an ugly approach to this which Lit makes easier, and so i pushed it back, but its still on the todo.

after that i should be able to have a more vanilla web dx.

Thanks!

https://www.reddit.com/r/cryptography/comments/1cint8h/what_are_your_thoughts_on_subtlecrypto_vs_wasm/

Tldr; I had that question before myself. It doesn’t seem to introduce any improvement over WebCrypto API which itself should have gone through peer-review and auditing. (“Choose your browser wisely” could be considered the advice here.)

I’d like to still look into wasm for several other details of the project, but it seems the built-in browser cryptography functions are performant and generally pretty good if used correctly.

no-installation. easy to get started without registration. can share with any device with a browser.

as a PWA, there is no need for you or your peer to already be part of the ecosystem.

the difference with my approach is that there is no registration or installation. as a webapp its easier for users to get started.

localsend is a good approach and my project isnt anywhere near the quality seen there (or several other examples). i hope to be able to create something competative in the space.

(i am investigating providing the app natively for various platforms for better reach, but the core requirement for my app is to have a modern browser)

the implementation is based on WebRTC which mandates encryption as part of the spec browsers have to confirm to.

in the app i have a concept of “contacts” in the app. when establishing a new contact, i use that webrtc channel to do a diffie-helman key-exchange (with RSA asymmetric keys used to exchange to AES symmetric keys). (im using vanilla os/browser cryptography tools as seen here: https://github.com/positive-intentions/cryptography/blob/staging/src/stories/components/Cryptography.tsx)

https://positive-intentions.com/docs/research/authentication

in my setup, all data that is being sent over webrtc, is already encrypted specific to a recipient. that encryption is a redundency over the WebRTC offering.

there is a performance hit with this redundency, but then allows me to have things like a way to verify public keys with a peer to prevent MITM vulnerbilities.

https://positive-intentions.com/blog/security-privacy-authentication

“file-transfer” as a capability is demonstrated in the chat app. the chat and file app are a GUI for peerjs-server.

thanks for your thoughts and reply!

i’ll try my best to answer, feel free to ask for clarity on anything i miss.

I would rather use Magic Wormhole if I have to have an intermediate server operated by somebody else.

completely understandable. its worth noting, what you see is a GUI for peerjs-server. under the hood its using peerjs. im working towards making it so that the backend is configurable to point to your own instance. it currently points to the public peerjs-server as a way for users to easily get started. this functionality is already available in the positive-intentions/chat app. (i will also be adding the ability to specify stun/turn servers).

ensuring that references to files are unforgeable and copyable

it seems i need to better understand the concept of capabilities as you mentioned for me to be able to reply on that. im not sure if this answers the particular concern, but i’ll try explain what is happening in a sequence.

1. user uses html input to select file from device. this requires the user to grant permissions.
2. file loaded into into os/browser/js memory-space.
3. this file is base64 encoded and sent to peer (base64 encoded to make the data serializable for network transfer)
4. peer recieves file in base64 which can then be converted back into a file.

i notice things like the file is sent incompressed without modifications. for file-types like images, i am able to do things like remove exif data, but this isnt implemented.

The terms of service are unacceptable and I won’t be trying out the product

this is fair. i dont know much about creating terms but i previsouly asked about if i can remove them entirely. i recieved strong feedback to keep it: https://www.reddit.com/r/startup/comments/1jerkyb/do_i_need_to_have_terms_and_conditions_in_an_app/

i think i should speak more broadly about the terms and condition. i was aiming to have something generic so i dont have to talk to a lawyer. i certainly dont have a lawyer on-call. i had the terms checked by a professional lawyer who i think advised correctly. the aim for the terms was to be something of a cover-all so that i dont have to worry about being personally liable. (its the same as used in the chat app.)

telemetry

im happy to consider flexibility on this. i was thinking that if i can check the domain isnt the one im using, then to disable telemetry. but the whole selfhosting-angle to this project needs a lot of consideration.

Just don’t collect user data

im not collecting user data. it seems a unique approach and i dont fully understand how to articulate it. its why i mention about it being zero-registration. its local-only storage for everything. there are no databases. its a GUI for peerjs-server. i selected to use nlevelanalytics because they provided something that i consider reasonably secure because i can interact with an api (in contrast, all other tools wanted me to add some remote script in the <head>, which is where i draw the line.). on nlevelanalytics the UI is faily basic. i see dots on a map. i dont think its enough to pinpoint individuals, but it give me county+city combo. im sure fingerprinting on network requests can still be done as much as any other tool. hopefully introducing ways to disable telemetry can be more relieving. but i dont particularly want to have a lawyer on-retainer frequently updating those terms.

somebody accused of misbehavior gets their local data exfiltrated too

there much to consider about a sensitive implementation for this. its an old post and i havent made much progress on it, but the question still stands: https://www.reddit.com/r/darknetplan/comments/16qw24o/on_my_decentralized_chat_app_i_want_some_kind_of/ … what is worded in the terms, is to allow flexibility on the implementation when it comes to implementing something for the purpose of “reporting users”. i actively discuss about the details and implementation of my project on reddit and details around exfiltrating data from users is clearly important to discuss.

ultimately, i often mention that the project is far from finished. your points are important to consider but also the tip-of-iceberg in all the things to consider. im no expert in anything and this is a learning experience to create something. i hope that by discussing it in posts like this, i can get to something more agreeable.

thanks for the clarity!

do you have any links to share about it?

By feature, do you mean “foss”? (Wondering if you’re replying to the correct thread)

If so, then it’s unfortunate I’m investigating this direction, it seems nessesary.

Otherwise feel free to let me know of a critical feature missing (if “foss” is not the feature you meant.)

No. When I type with “correct capitals”, it’s because I’m doing it from my phone. I otherwise generally might sound like chatgpt.

I’m no expert on the matter of licences either. I made the open source code some bsd licence because some of the dependencies called for it.

I created some code and made some open source. I don’t have to apply that to all my projects.