News for hackers. Stuff that matters.
Re:Dataset of classified screenshots from Tor hidden services by mdip on Thursday July 11, 2019 @11:58AM (#20412552)
I was reviewing the graph and was surprised at the rankings of things. Frankly, there's a lot more "not illegal" things going on in Tor than I would have expected. I expected there was a reasonable amount, but not representing so much of the total[0].

It's a double-edged sword in a way -- more people using the project for doing non-illegal things means perhaps the moniker "Dark" could be replaced with a less hostile word (Private? ...no...)[1].

Bad, in that, it speaks a bit to users making inconvenient choices because they perceive their privacy is threatened enough to warrant it.

I was glad to see that they scrubbed the content -- would have been one scary tarball to download, otherwise, and very illegal for them to post considering some of those categories. It makes me wonder how they avoided running afoul of the law just collecting/viewing it for the purpose of scrubbing it, but IANAL. It'd keep me from being involved on any project like that, though!

[0] I'm eye-balling the counts for a few categories in addition to the obvious "other-not-illegal", so it's probably not 100%, and I'm assuming US law (specifically WRT speech)

[1] I get "Dark" is meant as "invisible" and it's perhaps even a more accurate description since it's not completely "invisible" unless you take additional steps (and the software contains no known flaws), but sufficiently invisible for most adversaries. However, it's spoken in the same breath as "drug/theft markets" and is used more like "dark alley". Not sure there's a better term and I'm probably bike-shedding, anyway.

Re:Dataset of classified screenshots from Tor hidden services by VincentZ on Thursday July 11, 2019 @12:07PM (#20412654)
The graph displaying repartition was done only on a subset of all pictures. Not all of them are classified yet, and so, these stats were extracted from only a sample of this (quite big) dataset. Second thing is that these stats are computed BEFORE removal of all "bad pictures". Which means that what you have in the folder has not exactly the same ratio/statistics.

Concerning the "other-not-illegal", if the website was none of the other labels AND not illegal, it was usually tagged as "other-not-illegal". For instance, personal websites are tagged as "legitimate" and a Tor Wiki is tagged as "other-not-illegal" and "wiki". General information about Tor, websites allowing to do some calculation online (tools to hash things, calculator, ..) or online games (without money involved) are labelled as "other-not-illegal".

Anything related to finance - even if it can be legal - is not labelled as "other-not-illegal", for example.

Re:Dataset of classified screenshots from Tor hidden services by misterdoubt on Thursday July 11, 2019 @12:16PM (#20412773)
> It makes me wonder how they avoided running afoul of the law just collecting/viewing it for the purpose of scrubbing it, but IANAL. It'd keep me from being involved on any project like that, though!

CIRCL is the Computer Emergency Response Team of Luxembourg. They ̶a̶r̶e̶ work very closely with the law.

Re:Dataset of classified screenshots from Tor hidden services by LMYahooTFY on Thursday July 11, 2019 @02:28PM (#20414068)
Disclaimer: I'm very much a layperson here.

It wasn't obvious to me from the article how or if this dataset is statistically significant in some way in regards to total tor traffic?

I believe the tor project publishes anonymized data sets of tor usage, and if I recall correctly the web site handling the most traffic over tor is Facebook.

Re:Dataset of classified screenshots from Tor hidden services by dublinben on Thursday July 11, 2019 @03:20PM (#20414671)
If you look at the metrics for onion services traffic [0] and total network traffic [1] you'll see that this report is looking at the roughly 1% of tor traffic that is going to onion services. The overwhelming majority of traffic is to the regular, "clear" internet.

[0] https://metrics.torproject.org/hidserv-rend-relayed-cells.ht... [1] https://metrics.torproject.org/bandwidth-flags.html

Re:Dataset of classified screenshots from Tor hidden services by superkuh on Thursday July 11, 2019 @01:01PM (#20413207)
I couldn't agree more. I run a number of hidden services on Tor. That's because for every website I create it's a simple thing to make it a tor hidden service too. My ham radio and science hobby websites are in no way, "dark". I link to them (and vice versa) from my clearweb sites and host from my home connection/IP.

Plus on Tor you actually own your domain rather than lease it on the whim of some company easily pressured by political and social winds.

Re:Dataset of classified screenshots from Tor hidden services by mdip on Friday July 12, 2019 @11:14AM (#20421284)
Interesting. I've never thought to host my web sites on Tor. I don't have very many that are strictly "mine" but that'd be something interesting to do if only to understand a little more about Tor.
Re:Dataset of classified screenshots from Tor hidden services by mpfundstein on Thursday July 11, 2019 @01:29PM (#20413508)
How do you own it ?
Re:Dataset of classified screenshots from Tor hidden services by Diederich on Thursday July 11, 2019 @01:34PM (#20413539)
Your domain is a predictable derivative of your hidden node's key, which is randomly generated. It's cryptographically impossible for the same domain to be created more than once.
Re:Dataset of classified screenshots from Tor hidden services by superkuh on Thursday July 11, 2019 @06:10PM (#20416366)
Yup. Further, you can generate billions of keys until you find one where the public key matches some sequence you define beforehand. I brute forced one that starts with "superkuh". It only takes some tens of minutes to an hour or two on a GPU from 2010.

As long as you keep the private keys private you get traffic to that public key and "own" the domain.

Re:Dataset of classified screenshots from Tor hidden services by creatornator on Thursday July 11, 2019 @01:48PM (#20413702)
Not impossible, just unlikely. Collisions can exist in theory in any hashing algorithm (though it may be a 2^2048 large address space or something)
Re:Dataset of classified screenshots from Tor hidden services by Dylan16807 on Thursday July 11, 2019 @03:24PM (#20414732)
Which is what "cryptographically impossible" means. Even with all the computation on the planet for a million years you can't get a glimmer of a chance.
Re:Dataset of classified screenshots from Tor hidden services by hiccuphippo on Thursday July 11, 2019 @01:37PM (#20413577)
You generate a public/private keypair for your service. Your domain is a short hash of your public key. Whomever owns the private key, owns the domain name.