DELTA: Variational Disentangled Learning for Privacy-Preserving Data Reprogramming

Arun Vignesh Malarkkan, Haoyue Bai, Anjali Kaushik, Yanjie Fu · 2025

Read DELTA: Variational Disentangled Learning for Privacy-Preserving Data Reprogramming by Arun Vignesh Malarkkan, Haoyue Bai, Anjali Kaushik, Yanjie Fu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-00693.txt.

Description

Arun Vignesh Malarkkan leads a team of four authors on privacy in data preparation. Domain data often contain sensitive attributes about people. Rules such as HIPAA and GDPR govern much of that data. Clear feature engineering is also needed for trust. Yet standard engineering can leak privacy while it chases accuracy.

The team frames the task in exact terms. Features should be reshaped to raise target prediction. At the same time, they should lower prediction of sensitive traits. Their DELTA framework first uses reinforcement learning to find useful transforms. It then uses a variational sequence model with a split latent space. That split keeps utility signals apart from privacy signals. Across eight datasets, accuracy rises while leakage falls.

Catalog record

CREATORArun Vignesh Malarkkan, Haoyue Bai, Anjali Kaushik, Yanjie Fu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-00693

Browse the full library · About the Tower of Babel