5 ms·

There is absolutely no relation between the openness of a source code and its security I think that's overstating the case a tad. Take a look again at JGC's v

There is absolutely no relation between the openness of a source code and its security I think that's overstating the case a tad. Take a look again at JGC's very nice summary of the case in question: Worryingly, Haystack's only 'technical' detail is the following: "We use state-of-the-art elliptic curve cryptography to ensure that these communications cannot be read." Fair enough, but frankly that means nothing. They could be using AES, or RSA, or pretty much any good algorithm and I still wouldn't care. Two reasons: their implementation might be rubbish and enable attacks or their cryptography might be irrelevant because another technique (traffic analysis?) might make breaking Haystack possible. After all, all the Iranian government needs is a list of people running the software. In this specific case, having the source open would permit cryptography researchers (of sufficient skill and aptitude) to analyze whether or not Haystack is actually doing what it needs to do. Put another way: how do we know that Haystack isn't including a back door to capture all the information passing through their system, to sell on to the Iranian government for a hefty fee? We don't. I'm not arguing that open source is inherently more secure than closed source; I'm just saying that in some specific cases, there are clear benefits to having the source available for inspection.

jerf 16y ago
Elaborating on your point further: One of those "a little knowledge is a dangerous thing" errors that a security dilettante (like me!... though I'm at least far enough to know about this) can easily make is to assume that because you're using encryption, you're done. You're secure! Hooray! But of course it's not that simple. You have to consider your entire attack surface, and when dealing with a government that doesn't care if it disappears you for no good reason, there's one hell of an attack surface here. (Note: I neither know nor for the purposes of this post care if Iran is that cavalier, what matters is the existence of governments that are, somewhere, sometime, which I consider pretty likely.) They don't have to prove you've sent subversive stuff behind that encryption. They don't even have to prove you're using the subversive software. They just have to suspect it. Even if we assume the encrypted stuff is perfect, what other tells are there? Characteristic ports? Characteristic communication patterns? Characteristic headers? And even quick solutions to those problems, "oh, we make it look like HTTPS" can have problems of their own, ad infinitum. Are you doing HTTPS to sites that obviously don't serve a website? Are you trying to fake a website in a way easy to characterize? What will you do when the ISP straight-out bans websites on your home computer, making it impossible to mask your traffic that way? (And how suspicious is it for two home users to hit each other's "websites" every few seconds, anyhow?) These are all issues that a government won't have a problem answering, and there are yet more that they won't have trouble answering. The hostiles here have it easier because their threshold for deciding they have enough information to act is very low. The only way anybody should feel even remotely confident about this is if it is reviewed. (Whereupon it'll probably be discovered to be impossible, IMHO, but that's for reviewers to figure out.)
briansmith 16y ago
You can only verify that the software is doing what it is supposed to do by analyzing the resultant binary. And, analyzing the resultant binary can be done with or without the source code. In fact, many security researchers work exclusively on software (often Microsoft Windows) for which they will never be able to see the source code. http://cm.bell-labs.com/who/ken/trust.html http://cm.bell-labs.com/who/ken/trust.html

jerf 16y ago

Elaborating on your point further: One of those "a little knowledge is a dangerous thing" errors that a security dilettante (like me!... though I'm at least far enough to know about this) can easily make is to assume that because you're using encryption, you're done. You're secure! Hooray! But of course it's not that simple. You have to consider your entire attack surface, and when dealing with a government that doesn't care if it disappears you for no good reason, there's one hell of an attack surface here. (Note: I neither know nor for the purposes of this post care if Iran is that cavalier, what matters is the existence of governments that are, somewhere, sometime, which I consider pretty likely.) They don't have to prove you've sent subversive stuff behind that encryption. They don't even have to prove you're using the subversive software. They just have to suspect it. Even if we assume the encrypted stuff is perfect, what other tells are there? Characteristic ports? Characteristic communication patterns? Characteristic headers? And even quick solutions to those problems, "oh, we make it look like HTTPS" can have problems of their own, ad infinitum. Are you doing HTTPS to sites that obviously don't serve a website? Are you trying to fake a website in a way easy to characterize? What will you do when the ISP straight-out bans websites on your home computer, making it impossible to mask your traffic that way? (And how suspicious is it for two home users to hit each other's "websites" every few seconds, anyhow?) These are all issues that a government won't have a problem answering, and there are yet more that they won't have trouble answering. The hostiles here have it easier because their threshold for deciding they have enough information to act is very low. The only way anybody should feel even remotely confident about this is if it is reviewed. (Whereupon it'll probably be discovered to be impossible, IMHO, but that's for reviewers to figure out.)

briansmith 16y ago

You can only verify that the software is doing what it is supposed to do by analyzing the resultant binary. And, analyzing the resultant binary can be done with or without the source code. In fact, many security researchers work exclusively on software (often Microsoft Windows) for which they will never be able to see the source code. http://cm.bell-labs.com/who/ken/trust.html http://cm.bell-labs.com/who/ken/trust.html