본문 바로가기
  • Home

Edge-Cloud Pet Behavior Monitoring System Based on On-Device Encoding Anonymization

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2026, 31(3), pp.107~116
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : January 19, 2026
  • Accepted : March 5, 2026
  • Published : March 31, 2026

Hyuksoon Choi 1 JinHwan Yang 1 Nammee Moon 1

1호서대학교

Accredited

ABSTRACT

This paper proposes an adversarial learning-based privacy-preserving behavior monitoring system that fundamentally eliminates sensitive information at the edge device level. The proposed system adopts a Masked Autoencoder (MAE) as a backbone to effectively learn contextual features of behavior even in environments with limited labeled data. Specifically, we introduce a confusion loss into the encoding process to perform Min-Max optimization, which preserves the utility information required for behavior analysis while minimizing identity-specific information. Experimental results demonstrate that the proposed model maintains superior behavior classification accuracy compared to existing baseline models while significantly reducing identity re-identification F1 Score to a random guess level of 0.087(Behavior Classification), 0.052(Audio Detection) thereby validating the security and practicality of the system.

Citation status

* References for papers published after 2024 are currently being built.