AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

De-Anonymizing Text by Fingerprinting Language Generation

Zhen Sun, Roei Schuster, Vitaly Shmatikov

NeurIPS 2020 · Large language models › Language, safety and interpretability · approach: Controlled text generation · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use sequences of length N>=1900 instead of shorterfingerprint attack · reddit-sportsrecall: greater than 99\%Improved
remove 6% noisiest traces by noise level (ablation)fingerprint attack · reddit-sportsmeasurement error outliers: 6\%Improved
accounting for the 6% filtered tracesattack recall · reddit-sportsrecall: 99\% -> 93\%Hurt
apply attack to short chat posts (avg 9.5 words) instead of long forum postsfingerprint attack · Ubuntu Chatrecall: 93.4\%Mixed
restrict Silk Road users to those with variable nucleus sequencesfingerprint attack · Silk Road forumusers de-anonymized without false positives or negatives: 18 out of 41Improved
replace data-dependent top-p filtering with fixed-iteration constant-flow algorithmauto-completion sampler · loop execution timeslowdown: 1.15xHurt
fixed-iteration mitigationsample_sequence · 2700-word auto-completion simulationruntime: 0.3\%No clear effect
multiply measured iteration count by 100/1.1 to correct missed accessesFlush+Reload attacker · victim loop iterationsiteration estimate: 1.1\%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy