Aggregation-Aware Synthetic Text Generation for Authorship Privacy
August 25, 2026
AAST optimizes synthetic text generation at the bundle level rather than per-document to prevent cross-document authorship re-identification. The framework maintains semantic quality and sentiment alignment while reducing account-level linkability during stylometric and cross-genre attacks.
HOW THIS AFFECTS YOU
●
researcherYou can explore new methods for defending against cross-document stylometric profiling.
●
policyThis provides new technical frameworks for protecting user identity in multi-text datasets.