Hacker News Reader

Comment by kovac | Hacker News Reader

One loophole here is that statistical/machine learning analytics tools may still be able to correlate these posts as coming from the same writer.

We should probably also use a tool that just converts our posts into a generic style and remove any personal touches like the usual vocabulary, expressions, punctuation styles someone might use. Is there anything like that out there?

View on HN

It's exceedingly simple to correlate posts. I was able to use the stylo[1] R package to identify authors with just 3 tweets (requires testing against suspected authors), with 100% accuracy.

Robin Camille's master's thesis project[2] is one of the tools available to protect yourself against attacks like these. I had the idea to make something like this but learned that she beat me to it during my research phase.

[1] https://cran.r-project.org/web/packages/stylo/stylo.pdf

[2] https://github.com/robincamille/nondescript

I tried changing my writing style on my different personas and it is extremely difficult to keep up with. On one I'd never capitalize anything, another I would never use apostrophes, and then I had one that always used the wrong form of they're/their/there or to/too. Despite all those efforts, I could never consistently change my vocabulary and I did have someone call out a post of mine and link it to secondary accounts. I ended up ignoring the allegation and I never heard from the person again.

If there was some service to normalize, or better yet randomize writing styles, I'd love to know about them.

I don’t think you actually need to change the writing style, you just need to make sure that the corpus of writing for each and every identity is small. Then the error bounds on the probability that any particular piece of writing is by the same author as another piece of writing on the internet will be so wide as to effectively make such analysis useless.

Ideally, any site like reddit that allows pseudonymity would have a super lightweight way of creating multiple parallel new usernames for you to post under, with a way for you to see a coherent view of everything your multiple identities have posted, but where to everyone else, each identity looks like a separate user.

easily fixed: google translate your writing through a few languages before back to English

This is what I got with your comment, after going from English > Spanish > German > Vietnamese > Turkish > Chinese (Traditional) and finally back to English:

> Easy to fix: Google translates your text into several languages and then back to English

I'm actually a little surprised it was as close as it is. The form of the sentence has changed, but the meaning is still easy enough to figure out.

EDIT: Avoid adding Esperanto and Latin. That results in this:

> Easy to configure Google translates your text is in English and in other languages

1) Do not immediately use your real name to say funny things. Exit all social media, but "Congratulations baby!" Post a message like. This prevents you from meeting enemies who know your real name and reduces the chances of someone being you.

2) Start adopting a new pen name. Use different content on each site and use the password manager to stay in login. If you forget, don't worry. Treat all accounts as one-time accounts. that is not you. This is not "your brand". These are simply tools for putting on an anonymous mask and attending a walled garden party.

3) Do not mix the two online roles. Save your politics, jokes and personality in a pseudonym account. Keep boring and safe comments and dog pictures in your default account. Anonymous accounts do not talk about the weather in real life. Book such discussions only with family and friends. If you make a mistake, please delete your account and create a new one.

(English > Korean > Chinese > Hindi > English)

Reporter: how did you catch the suspect?

Police: it turned out Google Translate had a record of all their posts!

Reply on news.ycombinator.com