Hacker News Reader

Chaps.

I think it's time we just admitted that everything we type/tap/say into any device, regardless of how it's being wrapped up in transit, is absolutely and irrepairably insecure and we're not going to be able to fix that anytime soon.

Even with really smart crypto or some slick app; what makes you think that your messages aren't just being read by your os? Would you really notice? Why would you assume that the api this app talks to is trustworthy? There's no way to even tell if the code for the client on your device is the same as what's in that repo, as if it would make any kind of difference anyway since we can just load code from anywhere at runtime and you'd probably not even notice.

This whole thing is a fucking stageshow, I can only assume we're here to create an illusion that we still have any kind of security. Why? Is it to get some low hanging fruit/tech amateurs who will trust it and expose themselves? I don't know. But, HN, I don't know why WE aren't admitting this to ourselves or why we defend such obvious ruses.

Well, maybe some of you have signal stock I guess.

It's over, we lost, we have no security, we will probably never be able to reverse that situation, things are getting worse rapidly, we can't even know what any of our devices are really doing anymore (even the switches in your dc, the firmware on the UPS's, whatever), and apps like signal are obviously, to me, impossible to trust and I don't know why that's not obvious to anyone who has spent a moment looking at our situation..

Argh.

> I think it's time we just admitted that everything we type/tap/say into any device, regardless of how it's being wrapped up in transit, is absolutely and irrepairably insecure and we're not going to be able to fix that anytime soon.

We dont have to set the bar that high. Just lower it slightly and we can achieve great success.

Can we make the job of NSA and CIA harder? Not necessarily impossible, just to increase the costs they have to spend on whatever it is they are doing, much higher than it is now.

That would be nice.

We can do that with bitmessage, avoid Signal, use Telegram which means outsource to Russia/KGB - so NSA will for sure need to spend more resources if they want to access your "secure chat" on Telegram.

EDIT: but of course, as you say, the problem is socio systematic - any technical system which can or would be used to subvert the state is a threat to the state and will be neutrialized, by takeover, sabotage, lack of funding, the means the state has are enormous. The list we have of secure or safe systems are still zero. Only another state can go against the state.

For example, it was mandated that GSM had backdoors. Any system used for communication, especially mass/group communication, is not free and not allowed to exist. So simple is that. They even tried to go after the creators of secure systems, wasnt the PGP guy scrutinized?

I'm not sure there are many ways to make things harder when the complexity of the systems we have even in our pockets make them so very hard/impossible to even observe let alone protect.

There are too many places, even on a phone, where anything at all can be hiding and we don't really have any way of ever knowing what's there without stepping through the binaries.. That's even more difficult when we've got dalvik and VM's involved. How many people alive do you think could work out exactly what your mobile does in any one second within say, 6 months? ..

For all we know the thompson compiler backdoor really happens, and so how can you even trust the binaries you're producing?

Do you know anyone who built, for instance the openssl.so on their phone personally? Did that person build it on a machine they trust? How can they trust that machine? A single lib with a nefarious function on the few GB of binaries which ship with your phone is enough to completely negate any 'security' these sorts of apps could provide; even if they were trustworthy to begin with which they're obviously not.

It's a charade, assume everything you do is completely observed, regardless of crypto/apps you use.

Even if you have a fully audited and trustworthy (in your view) OS -- then what about IME? Who says that the CA that issued the certs you trust and talk to isn't also feeding them out to the gov or whatever? Who says your HDD isn't talking to the zigbee thermostat in your house and sending your rsa keys to the pentagon? Who says yo home wifi point isn't acting as a decrypting proxy?

We cant tell, and that's the point. There is no way to ever observe these levels complexity and verify security and we don't admit it.

We will never know if the code we are running on our various devices is doing anything extra or not, and so rationally, we should never trust it at all.

Not to say all sec is pointless, you don't want to be low hanging fruit, but above 'insta-pwn' levels of stupid there isn't much more available, even if you're told there is and we keep pretending like we have any control over what's going on anymore...

You're right, anything can be monitored, but we shouldn't take that as a reason that we should lay down and accept it. You should assume that any sufficiently motivated person/organization/state can get access to your information. But what we should absolutely do is make sure that state actors and companies have to demonstrate that they are motivated to take that information. That they have to devote resources to getting it. We shouldn't hand them broad access to everyone's information just because any one person's information is possible to obtain.

But regardless of that point, your other argument feels to me like a discussion against the problems of mono-culture: we should all endeavor to have diversity of implementation, if possible. The more we can work towards federated standards and the ability to quickly pivot when someone loses our trust, the more capable we are of defeating attempts to undermine that trust. If we only have 1 implementation of signal and everyone uses it, then all an "evil" party has to do is find flaws in that 1 implementation. If we have 2 implementations, then some users are protected. If we have 1,000 implementations, and they all inter-communicate, then we have some hope that only some users are affected by each compromise. I'm digging matrix.org as a federated protocol, and I'm hoping that it takes root because it has many clients; an open, federated protocol; and the ability for me to run my own server.

I would advocate that even though we're relatively mono-culture on linux, supporting it at least gives linux access to the source code which would allow other implementations (anyone else keeping a close eye on Redox [1]? I know I am!). Support open protocols! Support federated protocols! I'll advocate that signal is a great, easy-to-use system for secure communication, and that the ideas of Signal are percolating elsewhere (see "olm" [2] the implementation of Signal's crypto protocol for matrix). But I would like to see stabilization in Signal leading to federation.

[1] http://www.redox-os.org/

[2] https://matrix.org/git/olm/tree/

Indeed, another cost-increase on NSA and other evil enterprises invoice, is to misinform/disinform, let them monitor what is useless, waste resources finding the needle in the haystack.

Basically we are in Intelligence vs Counter Intelligence game with the state.

You suspect your phone or device is listening? I know that Facebook is listening/monitoring, so I let it hear. Yes, I did go to school in Zimbabwe and have my Zimbabwian friends confirm it, and I did visit it look the GPS data in the pictures show it.

The data that facebook has about me is worthless shit in disguise as marvelous worthy data, but they do sell it well, if everyone did this, it would be obvious their data is not worth the ads they sell, facebook would crumble like a stomped on paper plane.

I'm seeing the infosec community divide into several groups, and it's equal parts amusing and alarming:

* The "everything is fucked and will never get better" group, which I'd put your message into

* The "everything is fucked but you're not being targetted by Mossad" group - at least these people do threat modelling.

* People who are actively helping journalists in dangerous areas (see my first point)

I agree with lots of things in your message but I don't think it's helpful. Would you say we should drop back to HTTP because there are lots of ways to break HTTPS?

There are people doing fantastic work even on closed platforms like Android and iOS. People are looking in detail at what's running on it - mostly thanks to the jailbreaking and rooting communities. Everything will be OK, it's just going to take a lot longer than you'd like it to.

Maybe hiding our messages from Google is worthwhile even if we can't hide them from the NSA.

Yes indeed. And from criminals. While I don't want innocently become part of a government investigation merely because I communicated with someone who communicated with a suspect, it's well worth the effort to secure (or avoid) communication that criminals can intercept and subvert. That's the day to day danger to the day to day person, and that's why we need a secure by design internet.

However, I despair that a secure by design internet is impossible in this civilization, because it will be designed in an environment that includes "stakeholders" like Google, Facebook and the NSA via NIST. The only thing that you can do, beyond waiting for a white knight service, is button down your end, try to convince the other end to do the same, and use niche and custom built tools.

There is no secure future for the internet as we know it, or as it might come to be in the current civilization.

Chap (presumably),

> But, HN, I don't know why WE aren't admitting this to ourselves or why we defend such obvious ruses.

What exactly makes Signal an obvious ruse? Many of us don't admit that evading surveillance is impossible because we work full time in commercial surveillance. This is the currently dominant business model of the technology industry. Many of us have also worked in defense / government. At the very least, the fact that we have a non-commercial FOSS alternative to data-collection-traps, which is gaining traction, is fantastic.

> Well, maybe some of you have signal stock I guess.

Signal is run by a non-profit.

You just sound like you're spreading FUD.

> It's 2016 and our best crypto messenger options are worse than what we had 10 years ago when Skype was peer to peer, or Jabber with federation.

I strongly disagree, Skype has always been a black box. The usability of Signal's crypto is a lot better than it has ever been for Jabber.

> I can understand the reasons for not supporting federation, but I disagree.

Yet you don't address any of the problems with federation.

> What Open Whisper Systems and Wire need to do is open source the server components of their solutions, and try to remove the reliance on servers as much as possible.

https://github.com/WhisperSystems/Signal-Server

> Skype has always been a black box

I saw a BBC documentary in the mid-early 2000s (sorry, can't cite it right now) about law enforcement and intel having a really tough time with suspects using Skype, only being able to collect metadata and not call content due to its split routing protocols.

Now, there's nothing to say the whole service wasn't backdoored at some point, but it appeared to be a real concern, at least to LE. A few security audits at the time also gave it a very positive report (ditto for citation). In the post-MS sale era, of course there's zero trust. Prior to that though, things seemed different.

Matrix.org/riot.im has working federation, end-to-end cryptography, works without Google Play and is completely open source.

The encryption is audited, but not yet enabled by default, as there are still some rough edges. But the apps have tremendously improved over the last few weeks alone and I am very happy with it.

Thanks for posting this, I'm going to try and take a look over the christmas break.

Re: Google Play Services. Here's is the way I understand this to work, so please do correct me if I'm wrong.

Signal uses Google Play Services to notify me that I have an incoming message from Signal. The Signal app is then woken up and the signal app then retrieves my (encrypted) message. So, therefore, the only thing Google knows is that I received a Signal message, what time and if I have location services on, where I was at the time, yes?

On iOS, how is this any different? If I get a notification, iOS knows I got a notification from signal, when and where i was at if Location is turned on.

What am I missing?

Note: I _don't_ like that either system has that much meta data about my conversations available to them, but I don't see how one is better than the other, here.

The metadata is considered super important in modern surveillance. Your receipt can be correlated with a sender for example, over the course of many messages. Now they can start on a network analysis, start correlating with other people or events etc...

Oh, yes, I understand this. I don't understand how it is different for iOS than Android. Sorry if my initial post was not clear on that.

> So, therefore, the only thing Google knows is that I received a Signal message, what time and if I have location services on, where I was at the time, yes?

Considering that Signal compiles code from Google into their app, there’s an actual issue there: the Google code could just intercept and take control of any of the messages. And there’s a lot of code from Google in there that no one has audited yet.

What about on iOS? Couldn't the libraries that are used on iOS do the same?

> Re: Google Play Services. Here's is the way I understand this to work, so please do correct me if I'm wrong.

> Signal uses Google Play Services to notify me that I have an incoming message from Signal.

More importantly:

a. That dependency, along with other "restrictions", sets an artificially high barrier to actually using the product independently. This is presumably so that they can maintain the pretence of being "open source".

b. Keep track of application downloads and whatever else Google provides to developers.

Moreover, as the paper that someone has linked elsewhere says, if you are allowing Google services in your app (Google Cloud Messaging, to be exact), you're at the mercy of anyone with control of it. A relatively trivial attack via GCM, as the paper hints, would involve simply replacing your application with a backdoored version, and you'd be none the wiser. It is a massive attack surface that just cannot be ignored.

Matrix has the crypto and the federation. But of course its still not hiding metadata.

The Signal Server is open source, as far as I know.

> The Signal Server is open source, as far as I know.

Yes, except for the voice call component.

I just tried out Signal earlier this week with my wife and a friend.

The friend had a crash shortly after opening it on iOS, but it went OK after that.

You can't paste or drag in images directly on any client, as far as I can tell.

The "desktop app" is a Chrome app (not even Electron). Not loving that.

There were some hiccups with contact syncing.

It decided it should be responsible for alerting me to receiving ordinary phone calls (!?) on my iPhone, for some reason, which was annoying.

We abandoned the experiment pretty quickly :-(

[EDIT] it also lacked nice-to-haves one might be used to from other messengers, like auto-emojification.

I agree that decentralization and federation is an important step for security and privacy, reducing the dependence on a signal entity.

That said, based on OWS' writeup on the subpoena they received[1], the only user metadata they collect are the user's phone number(or a hash of it? for identification), that user's last connection date, and the dat etheir account was created.

[1] https://whispersystems.org/bigbrother/eastern-virginia-grand...

> It's 2016 and our best crypto messenger options are worse than what we had 10 years ago when Skype was peer to peer, or Jabber with federation.

Actually, Jabber with OTR is pretty solid. If need be, you can use throwaway addresses too.

> I can understand the reasons for not supporting federation, but I disagree.

I disagree and can't understand the reasons, which however I suspect to be rather shortsighted. Just imagine if email had not been what came to be called a federated system.

I don't know why SilenceIM isn't more popular. It's a fork of Signal's now-discontinued SMS encryption. Usernames are still telephone numbers (because SMS) and metadata is still leaked (to telcos, because SMS) but there's no reliance on any centralised servers or organisations, no requirement for Play services, and delivery is as reliable as regular SMS.

I don't know the real reason for Signal getting rid of SMS encryption but it's the only feature I'm interested in. I can get glorified internet messaging anywhere, but OTR encryption over SMS is a difficult problem due to the inability to conduct handshakes. We should not throw this work away.

What about WhatsApp, Telegram? I don't use them, but saying that there are no options today for secure, encrypted communication raises my eyebrow.

Telegram is a farce. They don't have end-to-end encryption by default, and their encryption protocol(MTProto) is home-made and basically a secret. They haven't opened it up to testing or open sourced it, so it's pretty useless at ensuring 'privacy'. Plus recently 15 million phone numbers linked to Telegram accounts were hacked because of an SMS verification loophole.

WhatsApp at least has end to end encryption- though that alone isn't enough, but it is not open source and is owned by Facebook. By default, it shares your number and your contacts' numbers with Facebook.

Neither Telegram nor WhatsApp are viable alternatives to anyone interested in privacy.

> though that alone isn't enough

Why? I'm not trying to nitpick, I just don't know much on the subject. Not being open source doesn't really concern me nor the fact that my number is known by Facebook. My telecom company knows it, my mother knows it and all my friends know it + also all the companies know it who are used by my friends.

> Neither Telegram nor WhatsApp are viable alternatives to anyone interested in privacy.

Are you suggesting that the application under discussion here is a viable alternative? If so, how?

>They haven't opened it up to testing or open sourced it

https://core.telegram.org/mtproto https://telegram.org/apps#source-code

Geniune question. For Wire: what is missing from what code they have available?

https://github.com/wireapp

Have they released the full server source yet? Last I checked they hadn't.

Yeh on researching a bit after this comment, I found exactly that. Server source code is still a "To be released" without a strong date. Other criticisms are generally around lack of full crypto audit, which as far as I'm aware (i.e, not very), is a criticism of most of the modern crypto-chat apps (with exception to OpenWhisper when it was TextSecure).

Servers are closed source, will open source in 2017.

Check out Matrix/Riot.