Privacy Preserving Machine Learning for Distributed IoT Security Systems
Fuente:
Zenodo
Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Recurso digital |
| Publicado: |
Zenodo
2022
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866901802053533696 |
|---|---|
| author | Sri Ramya Deevi |
| author_facet | Sri Ramya Deevi |
| contents | <p><span lang="EN-GB">The rapid expansion of Internet of Things (IoT) ecosystems has introduced significant challenges related to data security, privacy, and scalable threat detection. As IoT devices continuously generate sensitive contextual information, conventional centralized machine learning based security frameworks become increasingly impractical due to their reliance on data aggregation, vulnerability to single points of failure, and exposure to privacy risks. This article examines the application of privacy preserving machine learning (PPML) techniques within distributed IoT security architectures, emphasizing approaches capable of operating under stringent resource, bandwidth, and heterogeneity constraints. This paper survey core PPML methodologies federated learning, differential privacy, secure multi-party computation, homomorphic encryption, and trusted execution environments and evaluate how they support secure, collaborative anomaly detection without requiring raw data exchange. </span></p> <p><span lang="EN-GB">Building on this analysis, I propose a hybrid architecture that integrates lightweight on device threat detection with encrypted or obfuscated model updates, enabling continuous learning while maintaining strong privacy guarantees. Experimental simulations conducted on representative IoT datasets demonstrate that PPML frameworks can achieve detection performance comparable to centralized models while significantly reducing exposure of sensitive data and improving system resilience. The results highlight PPML as a promising foundation for next-generation IoT security, capable of balancing accuracy, efficiency, and confidentiality. This paper concludes by identifying open challenges related to scalability, adversarial robustness, and deployment feasibility, outlining directions for future research and real-world adoption.</span></p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18014676 |
| institution | Zenodo |
| language | |
| publishDate | 2022 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Privacy Preserving Machine Learning for Distributed IoT Security Systems Sri Ramya Deevi Privacy Preserving Machine Learning Differential Privacy Homomorphic Encryption Edge Computing Secure Data Aggregation <p><span lang="EN-GB">The rapid expansion of Internet of Things (IoT) ecosystems has introduced significant challenges related to data security, privacy, and scalable threat detection. As IoT devices continuously generate sensitive contextual information, conventional centralized machine learning based security frameworks become increasingly impractical due to their reliance on data aggregation, vulnerability to single points of failure, and exposure to privacy risks. This article examines the application of privacy preserving machine learning (PPML) techniques within distributed IoT security architectures, emphasizing approaches capable of operating under stringent resource, bandwidth, and heterogeneity constraints. This paper survey core PPML methodologies federated learning, differential privacy, secure multi-party computation, homomorphic encryption, and trusted execution environments and evaluate how they support secure, collaborative anomaly detection without requiring raw data exchange. </span></p> <p><span lang="EN-GB">Building on this analysis, I propose a hybrid architecture that integrates lightweight on device threat detection with encrypted or obfuscated model updates, enabling continuous learning while maintaining strong privacy guarantees. Experimental simulations conducted on representative IoT datasets demonstrate that PPML frameworks can achieve detection performance comparable to centralized models while significantly reducing exposure of sensitive data and improving system resilience. The results highlight PPML as a promising foundation for next-generation IoT security, capable of balancing accuracy, efficiency, and confidentiality. This paper concludes by identifying open challenges related to scalability, adversarial robustness, and deployment feasibility, outlining directions for future research and real-world adoption.</span></p> |
| title | Privacy Preserving Machine Learning for Distributed IoT Security Systems |
| topic | Privacy Preserving Machine Learning Differential Privacy Homomorphic Encryption Edge Computing Secure Data Aggregation |
| url | https://doi.org/10.5281/zenodo.18014676 |