On Tue, Oct 13, 2009 at 3:23 AM, VolodyA! V Anarhist <Volodya at whengendarmesleeps.org> wrote: > Evan Daniel ?????: >> >> On Mon, Oct 12, 2009 at 12:06 PM, VolodyA! V Anarhist >> <Volodya at whengendarmesleeps.org> wrote: >>> >>> Matthew Toseland ?????: >>>> >>>> On Monday 12 October 2009 09:54:05 VolodyA! V Anarhist wrote: >>>>> >>>>> >>>>> http://www.popsci.com/scitech/article/2009-09/gaydar-algorithm-outs-facebook-users >>>>> >>>>> A pair of MIT students claim that they have created an algorithm that >>>>> outs gay >>>>> members of Facebook by analyzing the sexual orientations of their >>>>> networks of >>>>> friends. >>>>> >>>>> The students first analyzed the networks of people who publicized their >>>>> sexual >>>>> orientation on Facebook. Turns out that statistically speaking, gay men >>>>> have >>>>> more gay friends than straight guys do. So then, they used an algorithm >>>>> to run >>>>> the stats on men who kept mum about their sexual orientation on the >>>>> site. Their >>>>> computer program was able to correctly identify 10 men whom the >>>>> students >>>>> personally knew to be gay in the real world but who hadn't shared that >>>>> fact on >>>>> Facebook. (The algorithm didn't work as well with women or with >>>>> bisexual >>>>> Facebookers.) >>>>> >>>>> [...] >>>>> >>>>> It'd be interesting if it's possible to make similar observations >>>>> without seeing >>>>> the whole map of the network, but only of the particular person. >>>>> Regardless, >>>>> this is yet another reason to dislike WoT. >>>> >>>> Only if WoT gives an option to publicly specify your sexual orientation! >>>> ;) >>> >>> True, thus WoT should never be used for anything which allows you to say >>> anything about yourself that can be associated with WoT identity. >>> >>>> And are you ranting against darknet here or only against chat/etc WoT >>>> apps? >>> >>> No, i was not ranting against darknet or only against chat/etc WoT apps. >>> >>> ? ? ? ? ? ? ? ? ? ? ? - Volodya >> >> How is this different from ranting about people engaging in the >> horribly insecure practice of replying to messages they find >> interesting? ?You can build the same basic graph structure, and >> analyze it in the same manner, merely from the messages. ?The WoT only >> makes it slightly more convenient, at worst. >> >> Evan Daniel >> _______________________________________________ >> Devl mailing list >> Devl at freenetproject.org >> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl > > Let me try to understand what you are saying. Basically you mean that a > reply to the message can be used in the same way to 'link' identities as > 'friends' as any social network would. So the premise would change from 'gay > people tend to have more gay friends' to 'gay people reply to more messages > of other gay people'. > > I must say that on the common-sense level i do agree with you in respect to > the fact that one can make such an observation, but i'd expect the > difference to be more than 'slight'. Especially for the people who make > their trust lists public (those who filter according to how much they agree > with the poster) responding to more of somebody's messages does not imply > that they will up that person's trust level.
Yeah, that's the basic idea. Except that (for starters) you replace "trust setting" with "number of messages replied to" or something. (I note that "friend" on FB is binary, so you just need an appropriate cutoff.) Whether there's a meaningful difference between "friend"/"trusted" and "replies to" is an interesting question. I suspect that there is a difference, but that the general approach is robust against it. But obviously that's only a suspicion; we can't draw a conclusion without evidence. IMHO this is merely an instance of the more general rule that Freenet and such can't protect you from revealing info about yourself. I suspect that if you only base your trustlist on the behavior of people on Freenet, rather than things like people you know offline and know what nick they're using, the risk is minimal. Remember also that the goal shouldn't be to prevent any information leakage; that's impossible. Rather, the goal should be to keep it no worse than (and ideally significantly better than) the leakage from things like word choice and who you reply to. Evan Daniel
