Deep Learning for Enhancing IoT Security using Multimodal Biometric Authentication

Abstract

Today, the Internet of Things (IoT) connects billions of electronic devices into multilateral computer networks to provide advanced and intelligent services. These networks enable numerous devices to communicate with each other for exchanging data and information with minimal human-to-machine interaction. This phenomenon increases the security issues and triggers the risks at a higher level in IoT systems compared with other computing systems. In order to maintain the security necessity when attacking the physical surface of the IoT system and its devices is a crucial and challenging task. On the other hand, implementing security mechanisms such as user authentication and access control for the IoT enabled ecosystems is essential to ensure the desired security of the IoT system devices. Usually, the security key may be stolen, forgotten, forged or duplicated by someone for misuse. The keys can be easily regenerated by intruders or men in the middle in traditional security environments. Today, biometric security is also becoming a more advanced and sophisticated alternative with technological advancements and is used widely in authentication systems. Technologically, only one biometric characteristic can be used in unimodal biometrics, which cannot be applied to ensure the high end security of IoT systems. In this research paper, we used biometric authentication to ensure the security of edge devices in the IoT environmental ecosystems. We also used the face images and fingerprint images as multimodal biometrics systems for authenticating users to secure IoT devices in an IoT environment. In the experimentation phase, we used a Pi-Camera module and a fingerprint sensor to capture biometric images. Then we used CNN algorithms for feature extraction and model development. As an activation function, the RELU function was used in model development, such as softmax for image classification, and Max-pooling for image dimensional reduction. This aided the model in speeding up the training process of the model. Finally, the experimental results demonstrate that the accuracy of the face image is 92% and the fingerprint image is 89%, which is a highly promising result to ensure the achievement of the desired objective of the research.

Keywords : Authentication, CNN, Deep Learning, Internet of Things, Multimodal Biometrics, Fingerprint