ORCID

Abstract

With the recent emergence of artificial intelligence (AI), edge users in industries and manufacturing have been extensively using AI-based services, which pose privacy and security risks to data. Distributed learning approaches deployed in the manufacturing industries help reduce data risks. However, traditional approaches cannot handle deep models as well as the scalability of edge-based internet of things (E-IoT) devices, especially in the manufacturing sector. Studies have proposed SplitFed learning (SFL) by combining split learning and the federated learning paradigm, but they fail to achieve an optimal trade-off between communication and computational limitations and are vulnerable to inversion attacks. We present a pruning-oriented security (POST) method that is designed around the SFL paradigm that not only helps in achieving a balance between communication and computation load for E-IoT devices, but also preserves the data and model privacy against inversion attacks. The POST leverages the concept of using a higher number of layers in the E-IoT devices, which restrains the attacker from reconstructing the outputs. Furthermore, the POST adds communication and computation constraints in the optimization function to reduce the overall cost of the method. The novel pruning method adopts the regularization and adversarial training approach to further improve the preservation of the privacy of intermediate features in the SFL paradigm. We conduct our experiments on publicly available datasets in real-world settings to illustrate the efficacy of the POST method in terms of preserving privacy while ensuring the best trade-off for communication and computation load.

Keywords

Data Privacy, Edge-based Internet of Things, Inversion Attacks, Split Learning, SplitFederated Learning

Publication Date

2026-01-01

Publication Title

IEEE Internet of Things Journal

Volume

13

Issue

3

First Page

4411

Last Page

4420

Deposit Date

2026-10-30

Funding

This work was supported in part by the Deanship of Graduate Studies and Scientific Research at the University of Bisha through the Fast-Track Research Support Program, in part by the European Union’s (EU’s) Horizon 2020 the Marie-Curie under Grant 847652, and in part by the Science Foundation Ireland under Grant 12/RC/2276_P2. Rdve 12 May 2025; visedre 20 July 2025, 25 August 2025, and 2 Nrvo 2025; accepted 10 Nrvo 2025. Date of publication 13 Nrvo 2025; date of current ersionv 26 January 2026. This orkw asw supported in part by the Deanship of Graduate Studies and Scientific Research at the Uyvn of Bisha through the Fa Tk Research Support Program, in part by the European Usn EU)’ Horizon 2020 the Marie-Curie under Grant 847652, and in part by the Science Foundation Ireland under Grant 12/RC/2276 P2. (Corresponding author: Abi Waqas.) Sunder Ali Kawh is with the School of Computing, Dublin City U,evn Dublin, D09 V209 Ireland, and also with ADAPT Centre, Dublin, D09 V209 Ireland (e-mail: sunderali. @wk .. Abi Ws is with the Tl National Institute, Uyvn College Cork, Cork, T12 R5CP Ireland, and also with the Tn Department, Mehran Uyvn of Engineering and T,e Jamshoro 76062, Pakistan (e-mail: Ab W@.. Mohammad Tz Quasim is with the Department of Computer Sci- ence and Artificial Intelligence, College of Computing and Information T,e Uyvn of Bisha, Bisha 67714, Saudi Arabia (e-mail: [email protected]). Kapal Dev is with the ADAPT Centre, Department of Computer Science and Munster Tl U,evn Cork, T12 P928 Ireland, also with the Department of Institute of Intelligent Systems, Uyvn of J,uro Jguro 2006, South Africa, and also with the Centre for Research Impact and Outcome, Institute of Engineering and T,e Chitkara U,evn Rajpura, Punjab 140401, India (e-mail: [email protected]). Digital Object Identifier 10.1109

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download


Share

COinS