AI can remove the mask of online anonymity. A recent new study, published on the arXiv platform, shows that advances in artificial intelligence may radically transform the way anonymous accounts work on the Internet.
The research team that conducted the study designed an automated system that mimics how a human investigator would work, but on a scale impossible for humans, and demonstrated that advanced linguistic models are able to identify users who post under pseudonyms by analyzing only the writing style and digital traces left in comments, reports TechExplore.
How the identification process works
The AI analyzes a user’s posting history on platforms like Reddit or Hacker News, extracts subtle peculiarities about language and behavior, then looks for matches in millions of public profiles, including professional sites like LinkedIn.
The model turns these bits of text – jokes, opinions, language tics, education or career information – into a mathematical ‘fingerprint’ of the user. Based on this, the system generates a list of possible real identities and calculates a confidence score for each match. If the level of certainty is low, the AI does not provide any conclusion to avoid errors.
Test results: Identification with 90% accuracy
To check the effectiveness of the method, the researchers tried to match almost a thousand LinkedIn accounts with the anonymous profiles of the same people on Hacker News.
Of course, all obvious data such as names, links or personal details were removed before testing.
The system was able to correctly identify users in up to 67% of cases, with an accuracy of 90%. By comparison, traditional, non-AI methods fell far short of these results.
The process is surprisingly cheap: between $1 and $4 for each identified account, the quoted source said.
Each seemingly mundane detail contributes to an ever clearer picture of the real identity
The researchers point out that “practical obscurity” that until now protected users with pseudonyms is no longer valid, warning that every comment, every seemingly mundane detail contributes to an ever clearer picture of the real identity, identifiable with artificial intelligence.
If such systems are perfected, they have every chance of becoming useful tools for areas such as cyber security or law enforcement.
At the same time, serious questions are being raised about the future of online anonymity and how much control users still have over their digital identity.
