Hacker News Reader

Comment by Zikes | Hacker News Reader

So the FBI knowingly attacked a civilian communications network, potentially causing a great deal of harm to the entire network in order to catch a few bad apples.

Like nuking a city because you're pretty sure you'll get a few bad guys.

View on HN

If the network can't ensure anonymity against bad actors, it was already fatally broken. If the consequence is improvements in Tor that break the attack, and the cost is a prosecution against someone breaking black-letter law, the result are a net-positive for civil liberties.

We can't ignore the potential collateral damage. Even the attempt, due to its inherently indiscriminate nature, is a violation of citizens' fourth amendment rights.

By your reasoning we should expect that they can kick in every door in a city looking for a suspect, then blame the doors for being too weak.

Can you be more specific about the collateral damage you're referring to? I don't understand the argument you're making. Perturbing Tor routing isn't in any way equivalent to kicking down doors. It seems that by your logic, even a title 3 wiretap "kicks down doors", since modern wiretaps happen on packet switched networks and involve filtering out other people's traffic.

If the collateral damage you're referring to is loss of trust in Tor, that's not damage: that's new information.

> According to a Tor blog post, someone during that period was infiltrating the network by offering new relays, then altering the traffic subtly so as to weaken Tor's anonymity protections. By attacking the system from within, they were able to trace traffic across the network, effectively following the server traffic back to their home IP.

This is network-wide. They did not and could not target just Silk Road 2 and its users, they sabotaged the anonymity of every user and service on Tor.

A title 3 wiretap requires trust in both the government and network operator, which users of Tor may be avoiding for entirely legitimate reasons.

I'm not seeing the moral distinction between title 3 wiretaps and Tor infiltration. It can't simply be that people on Tor have expressed an unwillingness to be tapped. Every rational actor has that preference in both scenarios.

Every rational actor has that preference, but not that expectation. I'm well aware that my ISP can eavesdrop on everything that happens over their network, and can extend that ability to any party they choose.

I agree that there is no moral distinction between a title 3 wiretap and Tor infiltration, however a title 3 wiretap is a passive listener while this Tor infiltration is not. The nature of the Tor infiltration caused anonymity to be stripped and readable by anyone aware of the flaw. They used resources unavailable to others to expose that information not only for themselves but to everyone.

The equivalent would be streaming a title 3 wiretap sans filter to everyone on the internet.

Perhaps you need to update your expectations? Merely expecting Tor to be secure does not make it so.

If you go through the exercise of building a mental model of exactly what a title 3 wiretap facility on a packet-switched telephony network looks like, and then carefully study the covert channel traffic confirmation attack the Tor team disclosed and everyone presumes the FBI is using, you'll see that the technical differences are not that great. It's certainly not as black-and-white as "passive" versus "active".

I'm not sure I follow the point about how the FBI could have done grave damage to everyone's privacy. It's (hypothetically, assuming this is how the FBI did it) the relay-early traffic confirmation vulnerability that did that. The FBI didn't create that vulnerability.

Wiretaps conducted by the FBI apply to installations in the US. Let's say I'm a dissident in my own (not the USA) country. An FBI wiretap does not compromise me wrt. my own government. Tampering with Tor on a large scale, however, has very much the potential to do so.

I'm not sure if this is what Zykes is referring to. But this particular attack on Tor is different from a wiretap because it made traffic readable by the whole network, not only by the attackers.

In what I think you mean by a perturbation attack, the attacker would deanonymize traffic by influence timing of packets on one end and observing the other end. Only the attacker learns anything. But in this attack, the hidden service directories found a clever way of broadcasting the name of the requested service, in plaintext, to the rest of the circuit. The attackers could read the message, but so could anyone else running a Tor relay.

Given that the message could have been trivially encrypted, that does seem like pointless collateral damage.

Tor is specifically designed to withstand attacks from resourceful adversaries who do not respect your fourth amendment rights.

The FBI putting it to the test provides us with valuable information both about the FBI and about Tor.

It didn't teach us anything we didn't already know about the FBI, but it also does not excuse their actions any more than if we were to attack any of their networks.

It didn't teach us anything we didn't already know about the FBI

Well, it taught us that the FBI is capable of carrying out a successful attack against Tor.

Maybe you knew that already - I didn't.

It's common knowledge that the FBI, DEA, etc have typically relied on the NSA for such exploits.

It is? Cite a credible source?

I'm thinking mainly of EFF, Bamford's books, and the Snowden releases. Are you claiming that parallel construction isn't SOP? Reaching way back, why do you think that the Weathermen got a pass in Chicago?

There are at least three levels of claim that can be made here.

1. The NSA passes along tips to FBI.

2. The NSA regularly passes along tips.

3. The FBI relies on NSA tips to do its job.

This doesn't make sense either. LEOs routinely force entry into buildings to execute search warrants. It is not hypocrisy that you'd be prosecuted for doing the same thing without the legal authority.

I contest the FBI's legal authority in this case due to the indiscriminate and damaging nature of their actions.

Warrants and wiretaps are narrow and selective, but this Tor intrusion potentially damaged the anonymity of everyone on the network.

You still haven't managed to make a coherent case for how this attack was any more indiscriminate than a wiretap.

As I understand this, the CMU team allegedly deanonymized SR2 in the course of their research, and then shared the information with the FBI. Once they had done that, they were allegedly forced to withdraw their Black Hat presentation. There's also the issue that the CMU team had NSA funding. I haven't seen any claims that the CMU team was initially focusing on SR2 and other "illegal" sites. But I wouldn't be surprised by that.

If that's what happened, it wasn't a wiretap. It was a tip.

CMU has had "NSA funding" for something like 20 years; it's the nation's software security "center of excellence", going back to CMU CERT, AKA CERT.

This is an especially silly bit of innuendo given that Tor is itself DoD funded.

The association of SR2 deanonymization with CMU's withdrawn Black Hat presentation is indeed speculative. But there's arguably more to it than innuendo. Time will tell.

It's well known that the DoD has funded Tor from the start. But it's at least decent of them to independently fund CMU to compromise it ;)

"Tor does not defend against a global passive adversary" -- TOR developers.

global passive adversary

Knowing that "global passive" might include the FBI is still valuable.

I feel like "global passive adversary" is kind of like a True Scotsman. There doesn't seem to be a fixed definition; rather we work backwards from random attacks and determine whether we think they were global adversary worthy or not, and if so then that makes the perpetrator a global adversary.

Could Lizard Squad have executed this attack? (I assume anybody with a botnet could start new Tor relays, so yes.) Is Lizard Squad a global adversary?

A global passive adversary is anyone who can execute a Sybil attack, basically. So the required size scales with the size of the network. A global passive adversary for a network with only ten nodes would only have to have five nodes itself.

The "global" adjective is just used, I think, because cryptographers presume a production deployment of the cryptosystems they discuss would be something like the Web: large enough (millions of nodes) to require globe-spanning resources (millions of other nodes owned by a single group) to execute the attack successfully.

Seen under that lens, neither Tor nor Bitcoin nor any other modern cryptosystem needs a "global" passive adversary to break it. Just a regular "passive adversary."

That's really interesting. So if there were multiple "global passive adversaries" then the network would become stronger and stronger? At least until one gives up and removes all their nodes at once.

Pretty much.

Imagine if China (used for population reasons) managed to send 300 million spies to the US to socially-engineer their way into all US citizens' personal lives. Now imagine India (again, for population reasons) simultaneously trying the same thing: now, one half of the time, the Chinese are just spying on "American citizens" who are really Indian spies, the Indians are just spying on Chinese spies, and one half of Americans go unmonitored.

It's sort of the same game-theoretic advantage you get from participating in a battle royale competition over participating in a 1v1 competition: for each new adversary you face, that adversary is also dragged down by all the other adversaries and becomes that much easier to deal with.

This really only applies specifically to Sybil attacks, though.

This was an attack by a partial active adversary.

Global = can view all network traffic. Partial = can view some portion of network traffic, but not all.

Active = willing/able to modify data as it transits the network. Passive = unable/unwilling to modify data as it transits the network.

The gold standard here would be breaking a specific user's anonymity without modifying the data, i.e. a passive attack by a partial adversary. The smaller the percentage of overall traffic that the system needs to observe, the better.

What about the citizens that dont have 'fourth amendment rights' just because they're not US citizens? As far as I know, Tor is not something that 'belongs' to US, nor 'exists' in US soil so it is very interesting what about the rest of the world.

I agree with your last sentence, as long as the black-letter law that is being violated is itself moral. I question the morality of many drug laws, however.

nuking kills. slowing down a network you are not forced to use? fine by my standards.

This isn't about the performance of the network. They could've brought it offline and that wouldn't have been as egregious an affront. They willfully subverted the network to invade the privacy and anonymity of the people using it. Regardless of the purpose, the potential damage to the people using it for any reason is too great to ignore. Political dissidents, whistleblowers, journalists, lots of people rely on Tor, and who knows how much of their information was exposed as a result.

If it's really that easy to subvert Tor then we should thank the FBI for showing us.

In the case of political dissidents in repressive regimes, compromising Tor can also kill.

There was a DEFCON talk about TOR where the speaker advised "All of my friends and contacts in Israel & the Middle East who didn't use TOR -- I no longer hear from them".

Billions are "collateral damage" in US government's "cyberwar". We even know now that NSA routinely captures botnets and "repurposes" them - anyone remember the Wikileaks DDoS? That one was quite obviously done by the NSA/FBI, but they may do it against other targets as well, in a much more subtle way or even blame it on someone else.

Over 80% of Tor hidden service visits are related to child pornography.* A 'few bad guys' is stretching things a bit. Additionally, they didn't nuke anything, Tor continued to function, albeit with a wide open security flaw that multiple actors besides the ones mentioned here were exploiting.

* http://www.wired.com/2014/12/80-percent-dark-web-visits-rela...

Under a very uncertain traffic measurement methodology which plausibly biases towards very evanescent .onion addresses.

And over 70% of emails are spam[1], so we should route every email we send through the FBI/NSA?

That other 20% of Tor usage may very well be political dissidents, whistleblowers, or average Joes that just don't want half the world watching everything they do and say online.

[1] http://www.kaspersky.com/about/news/spam/2013/Spam_in_Q2_201...

That's a terrible comparison.

Two points. First, the comparison of spam to child pornography is a bit... lopsided. Second, every email we send is made available to the FBI/NSA already.

"That other 20% of Tor usage may very well be political dissidents, whistleblowers, or average Joes that just don't want half the world watching everything they do and say online."

Or me, browsing my local government website to check when the next recycling pickup day is. Because I feel some kind of duty to do my bit to make the entirely innocent portion of the haystack bigger...

The title of this Wired article is somewhat over the top. It is about a (then) upcoming Gareth Owen's presentation on 31C3.

I just watched it [1] yesterday, and Gareth Owen himself says that this number is the count of a certain kind of hidden service request that should not be confused with "visits" or "visitors", for a number of reasons. The main ones I remember are:

- The specific kind of request measured is the first step towards connecting to the service, but may not always represent a complete connection that can be mapped to a individual

- Various anti-childporn organizations crawl the dark web constantly searching and indexing child pornography hidden services

[1] http://media.ccc.de/browse/congress/2014/31c3_-_6112_-_en_-_...

A more accurate analogy would compare it to DNS requests instead of visits. I don't know that there's any scientific research out there correlating percentages of DNS requests to visiting a website. That said, it's not an unreasonable assumption that the overwhelming majority of requests for any given domain hosting a web server are for actual users to fetch the content hosted there.

Over 90% of public roads service drunk drivers, drug dealers and most known and unknown criminals.

All users of such roads should be searchable at any given moment.

Your analogy would be more apt if the poster said '90% of nodes in the tor network routed child pornography'.

If it is true that 80% of hidden service visits are for child porn, then that would be more akin to saying 'X% of cars on the highways of america are drunk drivers or drug dealers'

This is unfunny because it is in practice true that any motorist can be searched at any time for any (or no) reason in the United States.

Tor is hardly a civilian network. It is quick and dirty secure communications for intelligence agents, developed by the Naval Research Laboratory. The civilians on tor are doing the useful job of adding routing complexity, which is probably why it was publicly released in the first place, as if it was only agents on there it would be useless for the job.

Reply on news.ycombinator.com