Live data from Hacker News

Ask HN: What are the implications of Hacker News alt accounts being revealed?

news.ycombinator.com

41–50 of 51 posts

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#42

If you're really paranoid/can realize the opsec ramifications, you'd note that submitting your username to a 3rd party site reveals a link of some sort between the submitted username and the IP the request came from.

How would it know which of the usernames that I searched are my account(s)?

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#43
I have no alts on this site (I just take the negative karma when I have something nice to say about crypto), but it still provided a list of 20 accounts, some with higher similarity scores than pg had (in the example).

Maybe I should be friends with these people then?

https://stylometry.net/user?username=cableshaft

EDIT: Did a quick cursory glance at the top five on my list, and despite pretty different subjects they talked about I can kind of see the stylistic similarities.

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#44

* It only works on accounts with 10k characters, alts tend to be for throwaway comments like your account * Deniability - dang has matches that are close but probably not him, you can just say it isn't you

None of the matches I saw for dang were me - but I think in this case the matcher was mostly picking up accounts that post a lot of links to past HN threads.

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#45

I have no alts on this site (I just take the negative karma when I have something nice to say about crypto), but it still provided a list of 20 accounts, some with higher similarity scores than pg had (in the example). Maybe I should be friends with these people then? https://stylometry.net/user?username=cableshaft EDIT: Did a quick cursory glance at the top five on my list, and despite pretty different subjects they…

My nearest neighbor is more distant than yours at .48. I suppose that makes my word choices more {weird,distinctive}. I'd like to see a list sorted by such a distance measure. Would voices with high distances have more distinct thought processes, or just a lesser grasp of the language? Would it be a measure of conceptual or cultural diversity?

I can't predict whether editing out highly diverse voices would increase or decrease signal to noise. Editing out low diversity voices would turn this into a stranger place.

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#46
post #25
post #2

I always assumed that some actors had this capability and acted accordingly. The existence of a publicly available tool merely democratizes this power. Anyone seeking complete and perpetual anonymity should have been practicing better opsec from the beginning.

This argument, which the creator of the tool also uses, strikes me as essentially, “I’m going to harm you to show you that there is potential for you to be harmed”. To make a crude analogy, what if I punched you in the face to demonstrate that you should consider learning kung fu? Or perhaps to indicate that, if a corporation wanted to, they could hire people to assault you and there would be little you could do abou…

People have been doing this kind of analysis for decades; all that happened today was that you got a demonstration of it, using a particularly simple approach.

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#47
post #6

>Presumably now people will routinely search for alt accounts of any HN commenter and bring what they find into the discussion. First, only obsessive or crazy people will do that, or maybe journalists. It'll show up occasionally in discussions, but I don't think it'll be 'routine', and I think most commenters will avoid subthreads with those sorts of shenanigans. But more importantly, these static analysis tools are…

Plausible deniability and then some.

The system doesn't predict alts, just gives a list of nearest matches and highlights the entries that are symmetrical within the number of matches given. I have no alts but still get the same number of "predictions" as sillysaurusx, who says he's had 10-ish.

So the accuracy of any given prediction is going to be by definition tiny at 19 per user. After all, most users have no or few alts in the first place. sillysaurusx is an extreme example, and even that leaves overall accuracy at 50-60%. Across all users I'd be shocked if it was better than 2 in 19. Probably it's < 1.

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#48

I finetuned T5, a text-to-text transformer, with a loss function based on the cosine similarity of any given text with statements made by individual HN users. It's like Dreambooth but for text. I have been posting under 40 different alt accounts that approximate the top users of this site textually and match them within an 80th percentile. AMA.

Are you serious? That is way cooler than the alt detector site. Depending on how much extra work you're willing to do I think TAILS is actually looking for a program to do this because they had some problems with Anonymouth.

Re: Ask HN: What are the implications of Hacker News alt accounts being revealed?

#49

Earlier quoted context omitted.

> Use newspeak Would only work if enough people did it. Otherwise all your accounts will stand out as being "the newspeak person"

> all your accounts will stand out as being "the newspeak person" Thanks, Obama (is talking in memes sustainable ?)

yarly.

tho (You) sneed to be perspicaciously circumspect in celebrating teh diversity of communities that u eBaum your MaiMais from. If Not, sekrit bsns liek whether ur gebaseerd 「kink memes」 are fr0m at least 100 years ago or $PRESENT_DAY_PRESENT_TIME will soon™ accidentally a pattern here, and you can rest assured that your FBI agent will have no doubts as to what kind of sussy you really are. )))

Post reply on HN