{

concerned_with_privacy: true,

online_usernames: ['kachapopopow', ...]

}

;)

As someone else mentioned, most likely you've been fingerprinted. But at least yes you can't be looked up directly, only if someone uses a databroker.

True, you don’t need to be impossible to be fingerprinted. If someone really wanted to track you manually using databrokers, be my guest :) But for most people here I think it’s about not being the easiest prey. And make the automatic and general algorithms hard to track you.

The "my bike lock is better than yours" strategy of personal information security :p

Good but temporary. Big tech has your browser fingerprint against that plus LLMs will probably be able to match it again by text using cosine similarly.

Maybe you use tails everywhere and run what you say through LLMs to rephrase. Might be OK then.

oh I've accepted that. luckily I have GDPR on my side.

With that use case you either dont have enough anonymity, or will forget the number of identities and leave a lot of traces, like admitting you are in the euro zone.

I don't care about that much if I did I would never use social media and sit behind a VPN 24/7.

For this you need to change more than just account. The way you write is your fingerprint. Concrete example with HN accounts was posted and tried several times.

Good thing I use AI to do all my writing.

This website analyzes speech to find your probable alternative names on HN. It’s probably easy to recoup a few more signals to find your name on other apps:

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

NB: Seems offline, but it was quite efficient !

thankfully my speech patterns naturally shift over time for who knows why, sometimes I start using commas too often. Sometimes I start my sentances with capitals, however, it's rare to see me put too much thought into things I write.

sometimes i just dont bother using the shift key at all

Unless you rotate IPs and browser fingerprints for each identity and alter your visited sites, usage patterns, typing speed and style, mouse movements etc. etc. the data brokers will be able to connect your ”inconsistent” identities.

Most attacks on people are done by those who just google your name and see what comes up so some very minor privacy work helps a lot. Its a lot of work to be completely safe but its very little work to be basically safe.

This was not about targeted attacks/doxxing, but systemic data gathering and enrichment.

The comment you responded talked about attacks/doxxing:

> I don't want to feel like someone can just dig up something I've said or shared 5 years ago and use that against me

"I have no idea who you are talking about leave me alone you freak."

if data is linked to me via some vague data that is based on similarities there's no world where that can be used as a trustworthy source or at least not putting doubt into the person who is using such data as often times it also links to a bunch of incorrectly assigned data. It's like trusting LLM's with everything they say.

Multimodal approaches are commonly cited as 89 to 97% accurate under experimental conditions. Certainly Google can be assumed to have much more data over a gazillion metrics, potentially making their accuracy close to 100%.

That makes a lot of sense, especially in a world where the internet never forgets