Deep Neural Networks for Dynamic Attribute based Encryption in IoT-Fog Environment

In a healthcare IoT based Fog System, malicious attacks may alter patient's critical health data, which may lead to severe repercussions. This makes it incumbent to adopt authentication mechanisms for preventing unauthorised access and implement data encryption to enhance the security in the system. This work presents an efficient learning integrated dynamic attribute based encryption mechanism by reducing encryption, decryption and communication costs associated in a Fog system with IoT devices and dynamic attribute updates. The Ciphertext Policy Attribute Based Encryption approach (CP-ABE) has been integrated with a Deep Neural Network model to use learning patterns related to attributes. This in turn reduces the communication cost incurred by the resource limited end devices for dynamic attribute updates. Further, an access control mechanism has been implemented by optimizing the system due to dynamic attributes and by analysing the updates in the access policy defined in CP-ABE. The Deep Neural Network has been trained using existing data sets and experimental results are presented by performing analysis on neural network parameters.