Z. Mahmood “Research and implementation of anomaly detection method using federated learning and blockchain in image datasets” doctoral dissertation defense

Author, Institution: Zeba Mahmood, Kaunas University of Technology

Science area, field of science: Natural Sciences, Informatics, N009

Scientific Supervisor: Prof. Dr. Vacius Jusas (Kaunas University of Technology, Natural Sciences, Informatics, N009)

Dissertation Defense Board of Informatics Science Field:
Prof. Dr. Hab. Rimantas Barauskas (Kaunas University of Technology, Natural Sciences, Informatics, N009) – chairperson
Senior Researcher Dr. Jolita Bernatavičienė (Vilnius University, Natural Sciences, Informatics, N009)
Prof. Dr. Gintaras Palubeckis (Kaunas University of Technology, Natural Sciences, Informatics, N009)
Prof. Dr. Agnė Paulauskaitė-Tarasevičienė (Kaunas University of Technology, Natural Sciences, Informatics, N009)
Prof. Dr. Filippo Sanfilippo (University of Agder, Norway, Natural Sciences, Informatics, N009)

Dissertation defense meeting will be at Rectorate Hall of Kaunas University of Technology (K. Donelaičio 73 – 402, Kaunas)

The doctoral dissertation is available at the library of Kaunas University of Technology (K. Donelaičio 20, Kaunas)

Annotation: Anomaly detection has recently emerged as a significant issue in a variety of fields, including picture In recent times, anomaly detection has become an important problem in many domains, including images datasets. It involves identifying patterns in data which deviates significantly from the normal behavior, indicating potential anomalies or outliers. Privacy concerns often arise when sharing sensitive data with third parties. Meanwhile, blockchain and federated learning technologies offer benefits in terms of data accessibility and privacy. Federated learning and blockchain are two technologies that can be used to address these concerns. A combination of these techniques provides a powerful tool for detecting anomalies in a secure and privacy-preserving manner, making it an increasingly popular approach in industries where data privacy and security is of utmost importance. In this work, we propose a framework for anomaly detection in images datasets using by federated learning and blockchain. To detect anomalies in image datasets by using federated learning (FL) and blockchain technology, a combination of supervised and unsupervised learning methods can be used. The approach involves multiple parties training a global model on their respective local datasets without sharing data directly, enabled by the FL technique. The blockchain technology ensures the integrity and transparency of the training process, thereby making it tamper-proof and providing a verifiable audit trail. The unsupervised learning method is used to identify anomalies that deviate from the normal pattern, while the supervised learning method is used to classify the detected anomalies. To enhance the privacy and security, zero-knowledge proof (ZKP) was used, which allows the parties to prove the validity of their local models without revealing their data. We describe the architecture of the system and provide experimental results demonstrating its effectiveness. Our results show that the proposed framework can achieve high accuracy while preserving privacy and security outperforms existing methods in terms of detection accuracy, while ensuring data privacy and security, making it suitable for various applications.