IFLBC: On the Edge Intelligence Using Federated Learning Blockchain Network

• Published in: 2020 IEEE 6th Intl Conference on Big Data Security on Cloud (BigDataSecurity), IEEE Intl Conference on High Performance and Smart Computing, (HPSC) and IEEE Intl Conference on Intelligent Data and Security (IDS) pp. 221 - 226
• Main Authors: Doku, Ronald, Rawat, Danda B.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-05-2020
• Subjects: Blockchains; Collaborative work; Computational modeling; Conferences; Data models; Distance measurement; Soft sensors
• DOI: 10.1109/BigDataSecurity-HPSC-IDS49724.2020.00047

Lately there has been an increase in the number of Machine Learning (ML) and Artificial Intelligence (AI) applications ranging from recommendation systems to face to speech recognition. At the helm of the advent of deep learning is the proliferation of data from diverse data sources ranging from Internet-of-Things (IoT) devices to self-driving automobiles. Tapping into this unlimited reservoir of information presents the problem of finding quality data out of a myriad of irrelevant ones, which to this day, has been a significant issue in data science with a direct ramification of this being the inability to generate quality ML models for useful predictive analysis. Edge computing has been deemed a solution to some of issues such as privacy, security, data silos and latency, as it ventures to bring cloud computing services closer to end-nodes. A new form of edge computing known as edge-AI attempts to bring ML, AI, and predictive analytics services closer to the data source (end devices). In this paper, we investigate an approach to bring edge-AI to end-nodes through a shared machine learning model powered by the blockchain technology and a federated learning framework called iFLBC edge. Our approach addresses the issue of the scarcity of relevant data by devising a mechanism known as the Proof of Common Interest (PoCI) to sieve out relevant data from irrelevant ones. The relevant data is trained on a model, which is then aggregated along with other models to generate a shared model that is stored on the blockchain. The aggregated model is downloaded by members of the network which they can utilize for the provision of edge intelligence to end-users. This way, AI can be more ubiquitous as members of the iFLBC network can provide intelligence services to end-users.