Listener Anonymizer: Camouflaging Play Logs to Preserve User's Demographic Anonymity
| Main Authors: | Kosetsu Tsukuda, Satoru Fukayama, Masataka Goto |
|---|---|
| Format: | Proceeding |
| Terbitan: |
ISMIR
, 2018
|
| Online Access: |
https://zenodo.org/record/1492509 |
Daftar Isi:
- When a user signs up with an online music service, she is often requested to register her demographic attributes such as age, gender, and nationality. Even if she does not input such information, it has been reported that user attributes can be predicted with high accuracy by using her play log. How can users enjoy music when using an online music service while preserving their demographic anonymity? To solve this problem, we propose a system called Listener Anonymizer. Listener Anonymizer monitors the user's play log. When it detects that her confidential attributes can be predicted, it selects songs that can decrease the prediction accuracy and recommends them to her. The user can camouflage her play logs by playing these songs to preserve her demographic anonymity. Since such songs do not always match her music taste, selecting as few songs as possible that can effectively anonymize her attributes is required. Listener Anonymizer realizes this by selecting songs based on feature ablation analysis. Our experimental results using Last.fm play logs showed that Listener Anonymizer was able to preserve anonymity with fewer songs than a method that randomly selected songs.