Semi-supervised and Unsupervised Privacy-Preserving Distributed Transfer Learning Approach in HAR Systems
- PDF / 1,625,074 Bytes
- 18 Pages / 439.37 x 666.142 pts Page_size
- 77 Downloads / 183 Views
Semi‑supervised and Unsupervised Privacy‑Preserving Distributed Transfer Learning Approach in HAR Systems Mina Hashemian1 · Farbod Razzazi1 · Houman Zarrabi2 · Mohammad Shahram Moin2 Accepted: 29 October 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract One of the challenges faced by machine learning in human activity recognition systems is the different distributions of the training and test samples. Transfer learning constitutes a solution to this problem. On the other hand, to perform transfer learning, it is necessary to have access to the original dataset. However, access to the dataset to implement the transfer learning algorithms results in a privacy breach. To deal with this challenge, this paper presents semi-supervised and unsupervised scenarios for privacy-preserving transfer learning in centralized and distributed manner. In the proposed distributed algorithms, it is not necessary to share the original data among the clients to implement the transfer learning algorithms. Instead, the transfer learning process can be fulfilled without having the original datasets. PPSETR and PPUSTR algorithms transfer the knowledge while preserving the privacy of the datasets on the client side. In contrast, PPDSETR and PPDUSTR algorithms provide the privacy protection of the distributed data on both the client and server sides. The proposed semi-supervised algorithms reduce the recognition error rate by 20.58% and the unsupervised algorithms decrease the recognition error rate by about 15.97% while these algorithms considerably preserve the privacy. Keywords Privacy · Semi-supervised · Unsupervised · Transfer learning · HAR
* Farbod Razzazi [email protected] Mina Hashemian [email protected] Houman Zarrabi [email protected] Mohammad Shahram Moin [email protected] 1
Department of Electrical and Computer Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran
2
ICT Research Institute (Iran Telecommunication Research Center), Tehran, Iran
13
Vol.:(0123456789)
M. Hashemian et al.
1 Introduction With the emergence of the Internet of Things (IoT), human activity recognition (HAR) systems have been taken into consideration due to providing context-aware information in various fields including medicine, personal living assistance, safety, military, and security [1–7]. The preliminary research has been focused on activity recognition using the motion sensors located in different parts of the body [8]. However, with the pervasiveness of the smartphones, the activity recognition using inertial sensors embedded in ubiquitous smartphones has attracted the attention of researchers [9–11]. Smartphones recognize an activity by either local or client-server approaches [12]. In the local approach, all steps of activity recognition, including data collection, processing, and classification, are carried out locally on the smartphone. In the server-client approach, sensing is fulfilled on a mobile phone as the client side. Then, the collected data is for
Data Loading...
