Live data from Hacker News

WhatsApp goes down for users globally

theguardian.com

231–234 of 234 posts

Re: WhatsApp goes down for users globally

#231
post #218
post #187

Earlier quoted context omitted.

> It uses a homebrew encryption scheme Please stop with that. At some point every encryption scheme was "homebrew" (or rather, new). MTProto 1.0 had flaws, which were addressed in 2.0. The latter has been independently formally verified to be secure[0]. > Telegram is the least secure. Against who? What is your threat model? Such absolutist blanket statements are useless by design. > of all the major messengers Which…

Your reasoning basically reduces to: I trust Telegram and I don't trust Meta. Security-wise WhatsApp is way more secure than Telegram because it uses the open-source Signal protocol [0]. [0] https://www.whatsapp.com/security/WhatsApp-Security-Whitepap...

Ultimately it always ends up being about trust at various levels, and how one can validate that trust, hence why I linked to "Reflections on trusting trust" by Ken Thompson.

That said, WhatsApp using the E2E Signal protocol is immaterial if they exfilter the locally stored private keys out to their servers, which is what the redaction of that line in the whitepaper we both linked (you directly, me via that Twitter link highlighting the diff) alludes to:

> At no time does the WhatsApp server has access to any of the client's private key.

Unless one checks every update of the client's code and the matching published reproducible binary output one can't be sure that keys (or any other data for that matter, since it has full access to anything decrypted) are not exfiltrated.

Now, that line deletion could be a subtle canary because law enforcement/state actor, or that could be Meta being nefarious following the cofounders departure, or anything in between. Either way I find it a worrisome signal that this specific line has been removed.

Re: WhatsApp goes down for users globally

#232
post #187

Earlier quoted context omitted.

> It uses a homebrew encryption scheme Please stop with that. At some point every encryption scheme was "homebrew" (or rather, new). MTProto 1.0 had flaws, which were addressed in 2.0. The latter has been independently formally verified to be secure[0]. > Telegram is the least secure. Against who? What is your threat model? Such absolutist blanket statements are useless by design. > of all the major messengers Which…

The other contender here is iMessage, which is likely more secure, but has such a terrible UX it’s not worth using. Their architecture is really poorly thought out and imitates SMS when it doesn’t need to. For example, if you travel internationally with a different SIM card, iMessage doesn’t allow your main phone number to be used for iMessage any more. Another terrible “feature” is that to have a complete history of…

> The other contender here is iMessage, which is likely more secure

Indeed, it is. I did not mention it because it's Apple ecosystem only, leaving a lot of people out. It appears to do many things very right, some lacking, and a few "wrong".

AIUI, with Messages backups in iCloud disabled, each device has its own key. Each message sent gets E2E encrypted with the key to each destination device and sent once for each device. So if the recipient has three devices, that's three encryptions and three messages sent. (That's how I recall someone describing it back in the day, I'm not sure today and I can't find the source of that anymore)

> Another terrible “feature” is that to have a complete history of your chats, you need to back up to iCloud.

The above means that a newly added device doesn't get access to the message history. This actually implements perfect forward secrecy! An attacker who manages to convince someone in some way to add a new device would a) be name to decrypt any old message intercepted and b) only be able to see new messages.

In that setup the only thing really lacking is being able to jointly check a contact (sender or recipient) key via a secondary channel and maybe TOFU it and displaying a warning when a contact key is added, changed, or revoked. You do get a warning for new devices added to your account but it could also apply to already added devices who unexpectedly get a new key.

Enabling backups in iCloud breaks perfect forward secrecy somewhat since the goal is being able to obtain the whole history, so an attacker managing to enroll a new device would presumably get the history. That said I also hear that this iCloud backup isn't zero-trust encrypted (technically it could be, think borg backup) but I'm really not sure about that.

> And uploading photos will quickly blow through your iCloud storage.

Not just that, it seems to do an absolutely terrible job at clearing the local cache, eating space like crazy and with no easy option to clear it: the settings app storage section is hopeless in that regard, Telegram's way of handling that manually plus the automated ones are muuuch more clear.

> For example, if you travel internationally with a different SIM card, iMessage doesn’t allow your main phone number to be used for iMessage any more.

I did not witness that when swapping SIM cards with new numbers: Messages popped up a dialog or something asking "keep using +XXXXXX" || "use +YYYYYY". I seem to recall I could even have both numbers for some time (IIRC there was a 2 week - or was it one month? - delay before a number is forcefully dropped out if you don't pop the SIM with that number back in). It was annoying the hell out of me as it was a short-lived number that was temporarily assigned while my real number was transferred between operators. My mistake though for tapping the wrong answer, but admittedly something there could use some improvement.

That was with local numbers though, the international story might be different? I would not expect that though as it would be truly an awful experience for international travelers that swap SIMs on non-multiSIM (eSIM+tray) iPhones.

And finally there's trust... it's completely closed and very hard to audit, but then again Apple owns the OS and hardware, so one could audit the app all they want, they have a much more potent vector for exfiltration.

Re: WhatsApp goes down for users globally

#233
post #104

Earlier quoted context omitted.

But nobody is forcing you to use matrix.org, and if you have the resources it's easy to run a separate instance with peering, so that you get both the benefits of being able to communicate with everyone and have a faster/locally controlled instance with whatever rules you like. We're keeping one alive with friends, not because we need it, but because we can and it's easy to do so. We can argue matrix has not the most…

I might not use matrix.org, but when 90% of my friends do (and currently that's closer to 100%), that doesn't do much for me... (But also: I don't have the resources to run a separate instance for myself, let alone for my friends.)

Perfectly understandable. If matrix took off a little more, we might see an increase in resources for the official homeserver, or see more peered homeservers becoming popular (as people today get a gmail account and expect to work with users on offic360 without questioning).

Meanwhile you also have the option to get a client with multi-account support. Again, not all clients have this, but this is again something that no other messenger does, or will ever allow purely for market reasons.

Re: WhatsApp goes down for users globally

#234
post #231
post #218

Earlier quoted context omitted.

Your reasoning basically reduces to: I trust Telegram and I don't trust Meta. Security-wise WhatsApp is way more secure than Telegram because it uses the open-source Signal protocol [0]. [0] https://www.whatsapp.com/security/WhatsApp-Security-Whitepap...

Ultimately it always ends up being about trust at various levels, and how one can validate that trust, hence why I linked to "Reflections on trusting trust" by Ken Thompson. That said, WhatsApp using the E2E Signal protocol is immaterial if they exfilter the locally stored private keys out to their servers, which is what the redaction of that line in the whitepaper we both linked (you directly, me via that Twitter li…

> That said, WhatsApp using the E2E Signal protocol is immaterial if they exfilter the locally stored private keys out to their servers

They don't do that.

Biggest issue with WA is the backup mechanism. You can encrypt it with a passphrase nowadays but AFAIK that's off by default, rendering E2E moot if either party backs it up to their Google Drive.

Post reply on HN