Federated Learning for Privacy-Preserving Malware Detection Across Distributed Enterprise Networks
Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023 · 20.12.2024 · 7 (01) · 460-474
DOI: 10.60087/jaigs.v7i01.517 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023, online erschienen am 20.12.2024. Verfasst von Sidhanta Panigrahy. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Malware detectors can improve when they learn from the telemetry of many organizations. Telemetry is confidential, commercially valuable, and often limited by contract and law. It can't be put together in one place, and when trained on local data, each enterprise is blind to the threat that everyone else has seen. We propose a reference architecture for privacy-preserving malware detection across distributed enterprise networks using cross-silo federated learning. We propose local training, update protection by clipping, differential privacy, and secure aggregation, robust aggregation against poisoning, validated and signed model rollout, and a governance layer. The article presents a threat model for seven classes of threats and compares them to a five-layer defense in depth, compares federated learning with four other techniques, analyses the communication cost of communication, and propose an evaluation protocol for security realism with time-aware splits, non-IID partitions, drift, privacy attacks, and poisoning This paper reviews recent papers that indicate that federated detectors can compete with centrally trained ones, and it identifies conditions where federated learning is beneficial and harmful. This article is a design and analysis contribution and reports no new experiments. It includes three tables and three figures to show the framework, and a roadmap of deployment and governance supports implementation.
Zitieren
APA
Panigrahy, S. (2024). Federated Learning for Privacy-Preserving Malware Detection Across Distributed Enterprise Networks. Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023, 7(01), 460-474. https://doi.org/10.60087/jaigs.v7i01.517
Vancouver
Panigrahy S. Federated Learning for Privacy-Preserving Malware Detection Across Distributed Enterprise Networks. Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023. 2024;7(01):460-474. doi:10.60087/jaigs.v7i01.517
BibTeX
@article{Panigrahy2024,
author = {Panigrahy, Sidhanta},
title = {Federated Learning for Privacy-Preserving Malware Detection Across Distributed Enterprise Networks},
journal = {Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023},
year = {2024},
volume = {7},
number = {01},
pages = {460--474},
doi = {10.60087/jaigs.v7i01.517}
}
EUVIMED