Twitter’s DM encryption is a modest security improvement over the status quo | Hacker News Reader

Twitter’s DM encryption is a modest security improvement over the status quo

platformer.news

146 pointsareoform70 comments

For anyone who's interested in a technical writeup of the implementation, I did one here: https://mjg59.dreamwidth.org/66791.html . There's code in the iOS app that would mitigate the MITM issue, but would also restrict users to only being able to read encrypted DMs on a single device, and it seems like they decided functionality was more important than security in this respect (which is not an unreasonable tradeoff, as long as it's well-communicated to users). Group conversations are also unsupported in the deployed codebase, so the concerns I had there may end up handled in some other way (though it's going to be difficult to manage that and also retain reasonable user experience). The perils of reviewing code before you can actually test the feature.

Overall I think my original conclusion still holds. If you're going to use Twitter DMs, turning on encryption is better than not doing so. But right now it's protecting against someone with opportunistic access to the Twitter prod databases, not an advanced attacker or someone with subpoena powers, and if you have to worry about that you should definitely use Signal instead.

(Edit: someone asked whether working on fruitflies would have been more relaxing than reverse engineering weird encryption, and I wrote this reply, but they'd deleted the comment. You get the reply anyway:

Reverse engineering fruitflies means reverse engineering something that's almost entirely undocumented, what documentation does exist is frequently wrong, that has almost 4 billion years of tech debt, that relies on self-modifying code, executed on a series of components that aren't fully deterministic, that can only be bootstrapped from existing fruitflies, that has no debugger, that had no functional specification, that has an instruction set where the behaviour of an instruction can depend on the three dimensional structure of the encoding of those instructions and so be influenced by instructions that are several thousand (if not million) instructions away in what looks like an entirely separate compilation unit, and if you change the temperature or humidity or any number of other apparently irrelevant factors will just behave entirely differently. And, if you fuck up, you suddenly have a room full of fruitflies.

So, yes, reverse engineering Java is definitely more relaxing.)

That last bit is how I sometimes feel about figuring out how the brain works compared to casually learning computer science stuff on the side. I love being a part of improving our knowledge of how the brain develops and sometimes goes wrong, but it's way easier to debug obscure source code than the brain.

I’d just add that reverse engineering fruit flies (realistically some component thereof) is actually much easier and relaxing than you suggest.

Because fruit flies are a model organism, they are therefore one of the most studied organisms in the world. We arguably know more about fruit flies than humans and try to extrapolate what we know about fruit flies to humans (with varying success).

I guess my point is that reverse engineering fruit flies is actually very well documented (in the biological literature, and even textbooks, very well established for decades now).

An unavoidable pain point when working with fruit flies is that there is no way to store them. A fruit fly is either alive, crawling around in a glass tube eating banana paste, or dead. Once it's dead, you can't do experiments on it, and you can't breed it.

Fruit flies are used because it is easy to do genetics on them. Genetics means causing random mutations, selecting for something, then breeding them true to make a mutant strain, then crossbreeding, making further mutants, and doing experiments on all the results.

Each of those mutant strains is part of your scientific 'capital', something you've invested in and want to extract returns from. You have to preserve it. And to preserve it, you need to transfer ('flip') flies of that strain to a fresh tube of banana paste every three weeks. For all of your strains. Without ever missing a flip.

So i'm not sure that working on fruit flies is relaxing, exactly. Maybe if you can afford a technician to look after them for you!

There is a lot of documentation on fruit flies. However, the share of fruit fly functionality that is documented is much less than 0.001%. This is fairly described as "undocumented".

How to prove Twitter can't tamper with the implementation on their server side?

Putting this out there. But I’ve audited a number of implementations of Signal and of close similar protocols proposed by some of the big players. Found a bug that broke Matrix e2e encryption[1]. Wrote the chapter on e2e encryption in the real world cryptography book including the only explanation of Signal in a book[2].

All I’m saying is that twitter, if you’re reading this, know that I have the blue checkmark ;) and that I’m available for pentesting your e2e encryption implementation!

[1]: https://matrix.org/hall-of-fame

[2]: https://www.manning.com/books/real-world-cryptography

Hey, David! Given the details we know right now about the implementation --- a single P-curve static-ephemeral key exchange for a long-lived GCM key --- would you sign off on that design? Is that how you'd implement encrypted messaging on a platform like Twitter? I mean, let's agree in advance that you might not do Signal Protocol given the expectations users would have. But is that how you'd do it?

Hey I’m not doing this for free! Hehe

Twitter has said this is an early version that shouldn't be trusted yet - https://twitter.com/elonmusk/status/1656570790039678976?s=20

It was brought up in the article. Elon only posted that after the security lead for the messaging product publicly tweeted completely wrong info about the status of the promised security audit.

Also, trust in the privacy of messages is the entire point of E2E encrypted DMs. It's a pretty dumb strategy to launch that feature and then walk it back with a "oops we're beta testing, don't trust that while privacy thing yet." So either Twitter/Elon has a dumb strategy, or they're (still) making wildly silly blunders as a result of self-inflicted wounds. Either way, pretty compelling arguments to never trust this feature even once they say you can.

Coming from the man behind the FSD “beta”, this is par for the course; but I still find it questionable to turn this sort of feature on in prod (for paying customers, no less) while simultaneously admitting it’s not trustworthy.

Perhaps it depends on what is meant by untrustworthy, but I don't see how it's a downgrade for users. Or are your concerns just PR-related?

It's an improvement over the status-quo. But given what was communicated before the feature was launched, it's very easy to believe that users might believe that the expectations around privacy and security are stronger than they actually are. Ideally you don't need to put any meaningful trust in the service provider, but in this case you do.

Well, I think if comms have been over-zealous then a correction is appropriate. But either way, this just seems now commentary on their PR chops, rather than anything fundamental about their technical implementation, so I'll colour myself unworried again.

It's better than the current product where it's readable in plaintext to Twitter employees, and to some outside contractors as well.

For those who already use Twitter DM's (which is a lot of people), this is an improvement.

If a security feature can't be trusted it shouldn't be deployed. Increasing trust is its primary objective.

>BigCivilEngineeringFirm says the new road bridge opened yesterday is an early version that shouldn't be trusted yet

There are many, many software developers. There are vanishingly few software engineers

The primary objective is not what you want it to be or what you think it should be. Elon - through all his career, has demonstrated that the primary objective is to overpromise loudly and underdeliver quietly - creating news, getting attention, being an edgy 14yo trapped in the body of an adult diaper wearer.

I see. FSD for encryption. Got it now. Thanks.

Fascinating that Musk continues to run Twitter in a "move fast and break things" mode.

When the entities that affect the company's existence the most i.e. regulators and advertisers are asking for more thoughtful, measured and safe changes.

And instead he's obsessed with adding end-user features which based on the woeful Twitter Blue subscription numbers seems like a dead-end strategy.

Twitter needs to improve its DM feature. It’s a nobrainer. They’ve been stuck forever on shitty DMs, they need to move faster.

Who are the people who actually GAF about e2e encryption in Twitter DMs though? Of all the places to have private conversations, that ain't it.

Twitter itself, because now it doesn't have to scan and moderate DMs for CSAM and other illegal activity. Implementing E2E encryption is a part of why they can run a skeleton crew moderation team. Users don't benefit from this change. In fact, more people will be harmed by Twitter's product with its inclusion.

source: NYT 2019 investigation into the link between the rise of social media and the spread of CSAM

https://www.nytimes.com/interactive/2019/11/09/us/internet-c...

Scanning 1:1 DM's is overly obtrusive. Apple iMessage and Facebook WhatsApp don't scan DM's either.

> Users don't benefit from this change.

Twitter blocked links to the New York Post in DM's over their Hunter Biden story. Clearly all users would benefit if they don't MITM 1:1 DM's.

There are entire communities (mostly schools and universities) that use Twitter Dm's almost entirely for communication.

I'm not sure why either.

It's a press release to drum up interest with the Musk-obsessed media though, and press coverage probably generates revenue.

Reporters, activists, whistleblowers, ... a lot of folks would benefit from improved security. But at some level everyone would. Importantly, Old Twitter blocked links in DM's to the New York post over their Hunter Biden story. You couldn't even share that story over DM.

Why ?

Twitter is built around public communication.

Why would I pay for encrypted private communication when I already use better existing solution for free ? And even if I pay it's useless unless others pay too (what a fail might have been interesting if it worked as soon as one party was verified).

With this feature they go from a shitty DM to a shitty sometime encrypted DM.

Wouldn't call this a nobrainer ...

> when I already use better existing solution for free ?

The good thing about Twitter DMs is that they don't expose your phone number which means you can - reasonably safely - converse with strangers.

Compeltely depends on your threat matrix I guess. At the moment (encrypted) Twitter DMs requirements payment, so that's a risk in some kind to safety. It's reasonably trivial to get an anonymous prepaid phone number.

There are many use cases. For example, a known government officials giving a tip to a journalist without giving them perpetual phone access.

There is a lot more friction involved in getting an anonymous prepaid number, and them proving your identity. On Twitter those issues are solved.

Twitter's a terrible choice for that use case, since the metadata isn't encrypted and you can't create an account without tying it to something else.

There is a time value in getting information out quick without having to create new accounts and re-verifying your identity.

> the metadata isn't encrypted

I wouldn't trust this if I was Snowden, but I would trust this if I was a corporate whistleblower and wanted to get information out to jousnalists as soon as possible.

Faster compared to who? Twitter wasn't build for dms, it was never the primary purpose, and even if it is a standard feature for all social apps, I highly doubt 99% of users care about e2e.

So? Facebook wasn’t built for DMs, can you imagine if they didn’t implement and perfected the feature? I personally use twitter DMs everyday and trying to move more convos to better messaging apps. They’re losing a lot of usage because of their lack of good messaging features.

I guess my point isn't that they shouldn't build a quality dm feature, it's that most people don't consider e2e part of that. The average person cares about ease of use/utility and that's about it. The people who do care more about privacy already use apps like Signal.

It's not "move fast and break things", he is just a horrible manager. Musk is a "big ideas guy", and he is good at pushing people to the breaking point to achieve those, but you have to keep him away from the important dials - like they do over at SpaceX and to some extent at Tesla.

Him micromanaging a business is a death knell. The guy is probably awake all the time from uppers, and you don't make good decisions where you are sleep-deprived.

He’s the richest person on earth and has a very solid track record. I’m betting he knows more than the average bitter commenter.

That's a losing bet since we actually have enough information about his knowledge beyond the simplistic heuristic "he's rich so he's right"

So you'll be willing to sell me calls in an Elon owned company then, yes?

This doesn't make sense whether you know he's failing or not. With that amount of money he can literally buy a temporary success when needed. That makes betting against his companies extremely risky - even if you're sure they'll collapse long term.

If Twitter could still be traded, he could announce his X app with some broken features, buy lots of ads, pay for influencers and reviews, and Twitter price would go up. And he's the kind of person who we know could totally do it.

For Twitter? Sure, it just wasn't possible anymore. And Tesla is still down 30 % as of now in the last 12 months, despite a 50% rally since January. I guess quite a few shortsellers made quite some money on that.

None of that is mitigated by the fact that Tesla is still up 718% since 2018, not everybody invested back then, nor is everybody, especially shorts and their counterparts in thier trades, long term investors.

These words you're using, I don't think they mean what you think they mean

Your question makes no sense without price information, and even then, that's way too broad a target to reflect the knowledge of a guy, where just listening is more than enough

a) He's not the richest person on earth.

b) His track record in general is not without faults.

c) His track record with Twitter is clear for everyone to see. Revenue has plummeted, Twitter Blue subscription numbers are just embarrassing and BlueSky is every day attracting more and more key content creators. Twitter is slowly but surely going to end up being nothing more than a Truth Social / Gab clone.

Revenue doesn't matter, profit does. Articles focusing on the former are being intentionally disingenuous. Twitter's employed headcount has fallen to to ~1800 from 7500, and their contractor headcount to ~1100 from 5500. And among the layoffs include most (all?) of the former executive positions, many of which drew quite absurd comp. For instance Vijaya Gadde's final comp was ~$17 million for a year.

That's going to be a dramatic decrease in costs. Their revenue since is going to be difficult if not impossible to fairly assess. Many (most?) of the largest advertisers like Apple/Amazon have returned, but likely with discounted enticements which, to my knowledge, are not public. So all one can do is speculate. I think the only fair conclusion is that it's hardly doomsday for Twitter, but it's also not quite time for a victory parade either.

I genuinely hope alternatives, especially decentralized ones, succeed. Competition improves everybody. But the reason Elon spent $40 billion on Twitter is not for the software, but for the users. Network effect is seemingly impossible to overcome, regardless of the merit of a site.

>He’s the richest person on earth

Mainly because of Tesla, which is overvalued also according to him.

>bitter commenter.

This fits for Elon too.

His track record at running Twitter has been abysmal so far.

Twitter works fine for me.

That's not incompatible with what I said. It mostly "works fine" for me too.

Works fine for me too. And am looking forward to all the new features they now have in the pipeline.

same

You can't squeeze blood from a turnip, and you can't make Twitter profitable. Twitter's problem is, and always has been, the business model. If you want to make money off ads then you have to keep the content acceptable to advertisers. Advertisers don't want their ads next to neo-Nazi propaganda on Twitter. Elon's approach to this is totally assbackwards if his goal is to make money.

Can you point to some examples of this propaganda?

Here is Musk himself trying to provide cover to one last week when he really had no reason to open his mouth about the shooting at all.

https://fortune.com/2023/05/10/elon-musk-twitter-texas-mall-...

Do you have a link to the tweets? I'm not reading through that

I am not your errand boy. I sent you the information you requested. You have all the details you need to research if you so desire.

I wasn't the one who insinuated that twitter was full of, and I quote, "neo-Nazi propaganda". And the comment history of the one who made that claim suggests an axe to grind.

Have you used Twitter?

yes?

Its idea that people covet a rich person just because they're rich is bizzare.

The post title is an… impressive spin on the actual headline.

Fake headline. Real headline is: "Why you can't trust Twitter's encrypted DMs"

I'd say one of the best features for private messaging is having disappearing messages. Then you lower your chances of leaking the messages to someone gaining illegitimate access right now, instead of leaking the database at any time in the future

Are you positive that the massages disappear from your device AND from the servers?

These companies notoriously do not delete anything ever. I am sure that even the things they are required to delete by law are removed, compressed and stashed somewhere.

As the reporting makes clear, it's kind of confusing who encrypted DMs are really for. Twitter finds itself in a position where it now answers to a man who is extremely vulnerable to state actors threatening his livelihood. In Turkey he became a fascist collaborator with virtually 0 push back. How much will he stand up for users of his platform if the CCP come knocking and want to unmask pesky protestors or users? Previously Twitter was blocked in China because they wouldn't follow the CCPs wishes. Well that was fine when twitter was independent they could stand behind their principles and China had no more leverage. Well what happens now that the CCP can shut down Tesla's Shanghai factory? Or even better, turn it over to a friendly local CEO? What happens when Elon Musk's line on "free speech is literally just what any local jurisdiction tells me" collides with US state bans on abortion, or LGTBQ rights? Sure would be awkward if the US suddenly had a lurch towards governors of right wing states introducing legislation to interfere with the operations of companies they don't like.

All the technical details of it are kind of moot, because fundamentally any attack on Twitter won't be technical, it'll be a practical application of pressure on Musk.

Elon covers it a bit in his interview with Tucker. Whole interview is interesting-ish, but he talks about who it's for around 23:00 on: https://www.youtube.com/watch?v=LCOe3a9EHJs

Elon's 7 months into a 17 year old company we finally have encryption.

But that's not fast enough because it's not perfect yet?

Do we think there's more dead wood Imposters that needs to be fired at Twitter to get progress even faster?

> Then they'd have the corresponding private key, and would be able to obtain the message encryption key.

This is technically wrong or badly worded. Is it just saying since you run their software on your device they could steal your private key? Same as WhatsApp or Signal? Not sure I get this point. If app developer turns evil and no one notices you are always stuffed.

> This is technically wrong or badly worded

Could you explain why?

> Is it just saying since you run their software on your device they could steal your private key?

No. It's saying that because there's no indication of which keys the conversation key is being sent to, if Twitter added an additional public key to a users's set of registered keys, anyone who initiated a DM with that user would have no way of knowing that they were sending a copy of the conversation key to Twitter. This doesn't involve any modification of the client code.

Theres no situation where Twitter develops and deploys E2E encryption.

Musk has already proven he is more likely to snitch on users for the government than the previous admin.

From there you can only extrapolate that this will be delivered with a government backdoor on the app side. If Musk actually manages to lure in whistleblowers to his platform with these antics those whistleblowers will have their asses hanging in the wind waiting for a government assistance request. Hopefully they arent as stupid as to trust musk.

It's actually plausible to audit the app for backdoor mechanisms, even without access to the source code. Right now there's no indication of one being present. I'm very very much not a fan of Elon, but in terms of the disparity between what's promised vs what's implemented right now everything points towards incompetence rather than malice.

Reply on news.ycombinator.com