Hacker News Reader

Comment by Vinnl | Hacker News Reader

What they've not addressed:

- Moxie's involvement with MobileCoin (could be as simple as saying: "for the same reasons Signal's interested in MobileCoin, moxie is as well").

- The concerns about regulatory risks/attention involved with a cryptocurrency integration.

- Why it involved hiding the server source code.

The idea of privacy-friendly transactions easily available within an app I'm already using is attractive, but I very much hope the obvious concerns won't play out in practice (which time will tell), and it would be good if they'd also addressed the above.

View on HN

Also: why Mobilecoin, which is as far from "an alternative payments infrastructure" as most other cash-grab cryptocurrency ?

It isn't known how much of it was premined, and from an early version of their whitepaper, the founders might have as much as 80% of the total supply of the coin. This isn't fine for a cryptocurrency which aims to be "an alternative payments infrastructure", and it isn't fine for Signal to try to promote them as such.

To play Signal's advocate, MobileCoin is one of the only cryptocurrencies that meets their requirements. The only other cryptocurrency that I know of that might come close is Grin.

The details of the distribution are lacking but as a beta it meets the experience they're going for - it's not as if they'd choose something that doesn't meet their requirements for the beta.

To clarify the only coins that meet their privacy requirements are ZCash, Monero, and MOB. The only subset of that that meets their transactional speed requirements is MOB.

At least AFAIK. There's a ton of alt coins so I might have missed something.

This seems extremely misleading to me.

Their "privacy" is strongly based on the confidentiality of SGX. If you're willing to trust SGX for your confidentiality and security, as they have, they could use any cryptocurrency in a highly private and instant manner.

E.g. via an implementation of the moneypot server in SGX: https://bitcointalk.org/index.php?topic=5302025.0

MOB is transparently a scheme to generate a windfall from shoving a pre-mined coin into a mass-market application. One could debate the ethics of that with a straight face, I suppose, but to argue that it is anything but is just dishonest.

My understanding is that SGX is only a layer. As in that it acts like ZCash or Monero and then has SGX as an additional layer. Even Signal uses SGX. But breaking SGX isn't going to break Signal. Also, my understanding is that these speculative attacks are still difficult to pull off in the first place. I mean everything is hackable, question is how easy it is to pull off.

The monero like part effectively runs on their servers inside SGX for performance reasons. If the SGX were completely insecure the users would have no privacy against the operators at all.

This is false, their privacy is similar to Monero (RingCT). SGX is only an added level of privacy.

No, that is smoke and mirrors.

The client fully identifies itself and its keys to the server, the wallet is entirely implemented server side. Without SGX there is no privacy what-so-ever.

Usage of ringct is incidental. It's like claiming that your documents are private when your operating system submits them all to the NSA because the NSA encrypts them after receiving them. :) (and, of course, we trust the NSA to behave honestly...)

You got sources for that being how the construction works? From various telegram groups with the devs in them , my understanding is 1) rings are constructed client side 2) uploaded to the server inside SGX. This has two major weaknesses 1) it doesn't specify how payment notification happens, which can be a massive privacy hole 2) if SGX is broken you are subject to all of the long term intersection attacks on Monero's/k-anonymity because the rings are now public.

This is a very limited threat model and worthy of serious scrutiny.

Wait, haven’t they published a white paper?

Also, _even if_ we steel-manned the design and imagine SGX is truly only used as a "defence-in-depth" layer on top of a ringct blockchain, using SGX to prove that SGX Lightning nodes weren't keeping transaction info would still be better privacy under many circumstances.

Lightning transactions are ephemeral: once the transaction happens, _no_ data about the transaction actually needs to be kept by anyone, let alone on a public blockchain. And only the peers directly involved in the route ever learn about the transaction directly.

Whereas with SGX, if SGX is broken in the future - and it will be - the statistical analysis on the underlying blockchain data is still possible. And unlike MobileCoin, this design wouldn't put SGX in any kind of trust or custodial position at all: if SGX failed, you'd just close the Lightning channels the same way any other Lightning implementation would, and open new channels with non-SGX peers.

I clearly misunderstood you but can you explain how you've come to that conclusion?

I'm trusting their whitepaper but it looks to me that fog provides similar security guarentees between MyMonero (shared view key) and other light wallets (computation on device).

I don't see how the privacy could be any worse than MyMonero even if SGX was compromised.

To your first point:

https://www.coindesk.com/signal-founder-may-have-been-more-t...

The thing that gets me the most though is not what they are doing, it's the way they are going about it.

The leader of a privacy focused app is now hiding secrets from everyone about the technology he convinced people was safe.

I've had to write off signal for this reason alone. I mean the only reason I ever used the app was trust.

> Moxie's involvement with MobileCoin

At best we know that he (claims to) have no coins. But that he was an advisor for them and focused on what he saw as a problem with crypto payments: speed. But yeah, there are still a lot of open questions here.

> The concerns about regulatory risks/attention involved with a cryptocurrency integration.

They addressed this before and the code reflects it. They whitelist country codes that allow it.

> Why it involved hiding the server source code.

This was weird and I'd like it addressed but probably not relevant to this particular topic/post.

> The idea of privacy-friendly transactions easily available within an app I'm already using is attractive,

Same. I mean it is a core element of the cyperpunk manifesto[0]. But I don't want to use a payment system that is volatile. If I have to choose to be beholden to services that collect my data or use a currency that is volatile, I'm going with the former every time. This is the question that I would have liked to be further addressed.

[0]https://www.activism.net/cypherpunk/manifesto.html

Even if he has no coins personally, that's a very narrow claim. Far more likely is that he has an interest in the company which happens to have pre-minted and owns most of the coins.

I mean, regardless of whether they addressed it before, to me this blog posts reads like their attempt to address all concerns raised in a single place, and these are all concerns I've seen raised. Thus, it seems to be like it would be very relevant to address them in this post, and would help restore some of the trust that some people already lost, and that others are at risk of losing.

> - Why it involved hiding the server source code.

My guess: to reduce the noise and speculation. Much better to deal with it now that there's something to be gained (feedback from beta testers and actual deployment). There was little to gain before, since the server accepts very few contributions anyway.

I wander enough on GH to see how much time big open-source project maintainers spend addressing toxic comments. Open-source is nice when you get relevant contributions and constructive feedback, but it's really not free, you have to deal with all the rest as well.

If you're not willing to be open with your source, don't market your software as open-source.

...there code always has and continues to be open source.

Is there some law that states a project must publish X code changes in Y timespan?

Well take it further – is is still open-source if you update the public repo once every 10 years?

For the vast majority of people, OSS in 2021 means having a public repo hosted somewhere that all commits are pushed to, or PRs are made to. Periodically updating a public repo with changes made to a private repo is not really what people have in mind.

The biggest example I can think of projects that worked this way was how WebKit and Android would publish code dumps with every release. You can't go longer than that: if you put out releases without updating the source you're not open source.

Were they updating the Server during the periods where the public repo wasn't receiving updates?

That's like requiring a developer to push their local changes hourly/daily for a task that's been committed to but not completed.

Yes, they were. There has been multiple "releases" too, with version numbers, which I assume went into production.

When you're using a product for which you can't get the source code, the product is not open source.

Commits: https://github.com/signalapp/Signal-Server/commits/master No push between: April 20, 2020 and April 6, 2021.

> Yes, they were. There has been multiple "releases" too, with version numbers, which I assume went into production.

You can't say affirmatively that yes they deployed but then say it's an assumption.

Either they did or they didn't.

If they didn't deploy the code to production prior to the commit then I see no problem with what they did.

I noted this in another comment, but if they _did not_ update the code in production, they ran with known vulnerabilities in the server side for many months: https://github.com/signalapp/Signal-Server/commit/3432529f9c...

There's no good option here.

Interestingly one of the comments in that thread refers to the server code as a reference implementation. This implies that Signal's server code isn't opensource, and that this code is just a reference implementation.

The readme.md on the repo doesn't really clarify one way or another on the matter.

I am 99% sure they did. I can't say affirmatively that anyone is running anything exactly, but they now claim they did.

> I am 99% sure they did.

Why?

> I can't say affirmatively that anyone is running anything exactly

Signal provides a means for reproducing Android builds so you can affirmatively determine what you're running.

https://signal.org/blog/reproducible-android/

> but they now claim they did.

Signal claims they shipped server updates without releasing code? Do you have a link to that?

When others were complaining about the lack of an update to server, they're complaining about the client and server APIs being out of sync. Reading through these comments, I don't get the impression that the server code published is what's actually running. In fact the following comment even refers to it as a reference implementation:

https://github.com/signalapp/Signal-Server/commit/3432529f9c...

At the end of the day, it doesn't really matter much what server code is running since the Clients manage encryption and the Server has no insight into the messages being exchanged.

And to take it even further, doesn't this bend the AGPLv3 licence to an extent?

If the code is entirely theirs they can do whatever they want including going all proprietary for all new releases tomorrow or moving to MIT.

Can they? First they'd have to never have accepted contributions, then if they ran an updated version on their server wouldn't they be in vialation of AGPL (not providing end users, in this case specifically those who use Signal an updated codebase).

If they have never accepted contributions, then they are the sole copyright holder, and thus can do whatever they want with the code. Including not providing the source code to users.

You can't sue the copyright holder for infringing their own copyright.

They have a CLA, the legal side of things is clear. They can use their code however they want, the licenses are only relevant for others. When you contribute, you basically license your code under a permissive license to Signal (because of the CLA).

In this case, they not only didn't update with changes, but didn't address specific concerns about known vulnerabilities in their outdated repo: https://github.com/signalapp/Signal-Server/commit/3432529f9c...

It is not about how often you update your project.

To be open source, you need to publish the source of applications you run. It's as simple as that: "we developed this app, and it's running on our servers, and the source can be found here".

They kept updating their server app for a year, without publishing their source. This is not open source. Signal server code was not open source for a year.

Isn't the allegation that they made changes to the deployed version of their production server that were not pushed to the GitHub project? (i.e., that they stopped updating the server repository for around a year, but not just implausible but contradicted by evidence that they never changed the server code for about a year?)

If so, then the code quite simply is not open source by any possible definition. I agree with you that a project that is dormant, or even receiving internal development that is still under review, can defensibly be called "open source" even if the development isn't happening in the open. (For instance, your favorite begrudgingly-GPL-compliant hardware vendor who releases .zip files every time they release a new product is releasing "open source" code.) But if the source code to the Signal server isn't even published, it can't be open.

I sympathize with this. On the other hand, you have to consider the big picture:

- the open-source argument was mostly meant for everyone to look at the sources and draw their conclusions regarding privacy and security

- these guarantees are mostly client-side (and you can't check what's running on the server anyway, even with the code)

It's fair to say that the server code was proprietary for one year. Most of their code was still open-source during that time, and most of it (now that they push to the server repo as well) is now.

I feel they would have taken less flak over this if their server code had never been open-source to begin with, or if they just released it now.

> My guess: to reduce the noise and speculation.

Why do you think more transparency would lead to noise and speculation? Exactly the opposite is the case. They could have gone the open route and maybe even gotten help from the community at implementing their payment feature.

But the way they did it they left everyone wondering whether their server is now spyware or what other reasons there could be for them to not update the public repo.

> Why do you think more transparency would lead to noise and speculation?

Because of the current situation: people argue about whether Moxie lost it, if Signal is still the right solution, etc.

If the feature had been implemented in the open, this debate would have taken place since the beginning, again with little benefit.

> maybe even gotten help from the community at implementing their payment feature.

Signal does not have an open development model. They're a small team, and most things beyond bug fixes are done by them. You can debate whether that's the most efficient approach, I'll just say it's an open question, and that's the model right now. So contributions to this brand new feature by the community would not have been possible.

> But the way they did it they left everyone wondering whether their server is now spyware

If people misunderstand that most of the security guarantees come from the clients, there's little an up-to-date repository on GitHub can do. I think the noise and complaining that has been happening for one year about the closed-source server is still negligible to what's happening now.

I would have preferred if they had kept it open, but I can definitely understand why they thought it was a better course of action.

Reply on news.ycombinator.com