Secured Data Transmission in IoT Using Deep Learning Technique for Data Encryption and Decryption Mechanism

Abstract

The IoT systems transfer highly sensitive data in a network and majority of the people needs IoT technology in real time applications like smart home, smart health care etc., A number of security algorithms exists to protect the IoT systems but with some time complexity. Deep learning is considered to be an efficient technique to analyze threats and respond to attacks and security incidents instantly and accurately. Conditional Generative Adversarial Network (CGAN) is one of the deep learning technique that protects data based on conditions created by generator and discriminator models. CGANs are useful for getting features of choice in generated data. This work use the CGAN feasibility to controllable the data encryption and decryption part in GAN network. This work solve the time complexity issues using Algebraic Matrix in Conditional GAN (AMCGAN) and Fully Homomorphic Encryption (FHE) algorithm. The advantage of using algebraic matrix is to reduce the time complexity and input complexity in cryptography process. It performs both addition and multiplication at the same time, and can compute any operation instantly. There are many encryption techniques used to encrypt the data but the same time they have more time to decrypt the result. Because the mathematical evaluation are complex to derived in encryption part and otherwise. So, this work considered to address the time complexity problem by solving the easiest mathematical derivations in encryption and decryption part. Also we noticed that the fully homomorphic encryption algorithm have less encryption time compared to Chaotic Algorithm and it has minimal time complexity than existing algorithms such as RSA algorithm.