Haystack project responds to 'security concerns', looks like it's falling apart

http://blog.jgc.org/2010/09/haystack-project-responds-to-security.html

by thomas11 • 16 years ago

44 23 16 years ago

blog.jgc.org

j2d2j2d22• 16 years ago

I'm confused by some points.

1) Why were they writing their own system to do this? Haystack sounds similar to Tor, which is probably more tested.

2) How is hype being allowed to succeed in decisions made to address very real causes? The glamour of hacker culture appears to have won, over the science of hacking.

jgrahamc• 16 years ago

I think the media liked the 'Austin Heap' persona. A combination of that and a hot topic like Iran meant that many outlets decided to put normal journalism aside and report what looks like a non-story.

Hell, The Guardian made him Innovator of the Year: http://www.guardian.co.uk/media/2010/mar/29/austin-heap-mega...

tptacek6• 16 years ago

The really unfortunate thing about the Haystack debacle is that it's going to be framed as a problem with Austin Heap and his team --- as if a better team might have been able to pull something like Haystack off --- when in fact the problem is with the fundamental economics of pitting private activist groups against foreign governments on engineering problems.

JulianMorrison• 16 years ago

Haystack and Tor do different stuff.

Tor conceals your origin from the endpoint (web server etc).

Haystack conceals the fact you are communicating and about what, from a man in the middle, by pretending to be innocuous web requests.

j2d2j2d22• 16 years ago

Ah, thank you for the clarification. The Tor project's page says the same.

Tor anonymizes the origin of your traffic, and it encrypts everything between you and the Tor network and everything inside the Tor network, but it can't encrypt your traffic between the Tor network and its final destination. If you are communicating sensitive information, you should use as much care as you would on the normal scary Internet — use HTTPS or other end-to-end encryption and authentication.

http://www.torproject.org/download.html.en#Warning

mcantelon• 16 years ago

tptacek6• 16 years ago

Here we have a painful example of the biggest problem with circumvention tools: the people writing them don't have the resources to compete with their adversaries.

It's easy to assume that knowing how to code with sockets and call into a crypto library makes you more clueful than the people running "the great firewall" or whatever it is Iran has. They're trying to censor the Internet, how smart could they be?

Well, they're smart enough to pay someone a million dollars to break a circumvention tool without giving it a second thought. The Tor project, which actually has some software security talent, couldn't clear the "adversary with $1MM" bar (they screwed up Diffie-Hellman). And Tor has an "easy" problem; stealth circumvention is much harder.

Get circumvention at all wrong and you achieve the opposite of what the tool is intended for: you put a big red flag on people breaking their local laws. The risk/reward structure here is totally broken, even before you consider how likely it is that everyone's machines in the country you're trying to "help" are already rootkitted.

Don't build circumvention tools.

aristus• 16 years ago

Do you have an alternative idea? Or are we doomed to police-statism? Serious question.

tptacek6• 16 years ago

I don't think it matters if I have an alternative idea. The world is under no obligation to be equitable. Meanwhile: "best effort" isn't good enough when the penalty for failure is someone being hung from a construction crane.

I think a state-sponsored circumvention tool, backed by one or more western governments, would --- if procured properly, which would never happen, and instead would involve Lockheed or SAIC --- stand enough of a chance against a foreign adversary to be worth considering. Short of that, you run right into a very simple wall: once your tool starts to matter, Iran will happily spend many millions of dollars to have it turned.

jerf• 16 years ago

I would say we are not doomed to police-statism, because there is a countervailing force: An open network is more valuable than a closed network. There are forces that will work to keep the network open so they can tap that value, though they may face a coordination problem. Google would be one example of a company that, for all its many missteps and concerning moves, is broadly speaking fighting for a more open net rather than a more closed net. (I'm not saying they've been 100% successful or 100% aligned to openness, but net-net I've been reasonably impressed with their vision on this point; and again, by "reasonably impressed" I do not mean "in love with".) Even a government may be convinced that the more valuable open network is in their best interests, as it represents a larger tax base to work from. Perhaps we should be encouraging Internet taxes now, instead of fighting them!

But the police state won't be defeated by broad-scale usage of some technical tool. Small-scale usage by a small core group of hackers who constantly adapt (and, frankly, constantly suffer attrition by the police and face the non-zero chance of making one mistake that allows the police to catch them all) might be possible, but if we're going to defeat a police state it will have to be on something other than a purely technical level like that.

mindslight3• 16 years ago

> Don't build circumvention tools.

Could the entirety of your arguments against cryptosystem design not be applied to SSL in its infancy? Aren't the resources of the attackers always larger than the resources of the developers?

There's clearly a large gap between what currently accepted protocols can do, and the capabilities of zero knowledge proofs and secure multiparty computation. It's certainly smart to reuse existing blocks rather than recreating their errors, but that doesn't mean all crypto protocol design work is antiquated.

(although please don't think I'm saying distributing Haystack to users in repressive regimes was anything but criminal malpractice. Even TOR says to not rely on it for strong anonymity)

tptacek6• 16 years ago

If SSL was being sold as a tool to allow people to visit censored Internet sites from ISPs controlled by a government that still kills people by burying them in a sand pit and hurling giant rocks at their head? Yes. It'd be the same argument.

Since I never argued that all crypto protocol design work was "antiquated", I'm not sure how to respond to that. What I've said a bunch of times is: it is prohibitively difficult to get stealth circumvention tools right on the first try, and the adversary being circumvented could pay the couple million it would take to turn any such tool, without even noticing the cost.

eli• 16 years ago

On The Media just aired a piece titled "Is Haystack Too Good to be True?"

http://onthemedia.org/transcripts/2010/09/10/05

benatkin• 16 years ago

Is this the Haystack that forced 37signals to change Haystack's name to Sortfolio?

malkia• 16 years ago

It looks like there few other projects with the same name.

bmelton• 16 years ago

I hadn't heard anything about this until just now, but man, am I really glad that this isn't http://haystacksearch.org/, as I thought they were talking about.

On the flipside, while the goals of the project seem noble enough, I can't help but wonder why they didn't just try to value-add to something like Tor.

Regardless, I'd much rather they stop distributing while they get things sorted out than offer a false sense of security to their users who could potentially be using Haystack in ways that, if caught, might put their lives/careers in jeopardy.

sweis• 16 years ago

Some background on this: Jacob Appelbaum obtained a copy of Haystack and found serious vulnerabilities that could put users at risk. He convinced Haystack to immediately suspend operations.

From what I understand, there was a diagnostics mode that allowed an attacker to identify running copies of the program -- essentially the exact opposite of what it's supposed to do. Even with Haystack's proxy servers shut down, there may still be a risk to any user who has it installed.

This could have easily been prevented. Many people with security backgrounds, including myself, contacted Austin Heap and asked for technical information. He declined. Instead, they went ahead and distributed it to vulnerable live testers.

Haystack project responds to 'security concerns', looks like it's falling apart

Related Stories