An Explorative Analysis of IoT Security on Federated Intelligent Networks

This paper describes the use of deep learning techniques in order to analyze the data generated by IoT devices and determine whether the data was generated as a part of the normal operation of that device or was it generated by a malicious node. The malicious node could cause attacks like denial-of-service attack or a brute force attack to bypass the password authentication of the target device. Deep learning is suggested to be an effective way of providing security to the devices that participate in an IoT network. This paper describes federated learning techniques which are utilized since the IoT devices tend to have less processing power sufficient for the normal operation of the device while conserving the rest in order to minimize the cost. Certain vulnerabilities arise in the system due to the use of federated learning techniques namely lack of incentive to individual nodes and poisoning of the global model. We propose a blockchain based solution to provide incentive to the individual nodes in terms of credits. The incentive is in proportion to the amount of contribution provided by the given node. The resultant updates generated by compute nodes are verified based on the impact they will have on the global model before incorporating them into it.