AUTHENTICATION IN IOTS AND SMART GRIDS VIA FEDERATED LEARNING AND BLOCKCHAIN

---

Tezin Türü: Doktora

Tezin Yürütüldüğü Kurum: İstanbul Sabahattin Zaim Üniversitesi, Fen Bilimleri Enstitüsü, Bilgisayar Mühendisliği, Türkiye

Tezin Onay Tarihi: 2026

Tezin Dili: İngilizce

Öğrenci: Ahmad Awni Jawdat ALI

Asıl Danışman (Eş Danışmanlı Tezler İçin): Mohammed J.M. Wadı

Eş Danışman: Vısam Elmasry

Özet:

Data sensitivity, server centralization, and device identity verification are three critical

security challenges that Internet of Thing (IoT) and smart grid (SG) systems will have

to face. This thesis proposes a novel integrated solution with high security and

intelligence without compromising efficiency, which can be achieved through the

integration of blockchain (BC), federated learning (FL) and authentication

technologies (ATH).

The analysis utilised a three phase experimental method. The first was performed by

analyzing four FL models and five aggregation techniques (AT). The results indicated

the extent to which the FedAvg and FedProx algorithms were better. The second step

was the integration of blockchain technology. While they used a modest increase in

computation time and reasonable power consumption, the outcomes showed

substantial gains in model robustness to attack. Step Three included adding multifactor

authentication (MFA). For the final integrated system accuracy was 55.54%

under a 30% attack with training time reduced from that of base system (52.82%).

This study also presents an integrated system for understanding algorithm behaviour,

establishes a three stage evaluation method and provides practical evidence in order to

improve the efficiency of time and energy. Combining these three technologies

improves both security and efficiency simultaneously, thus opening the possibility of

decentralized FL in highly sensitive applications.