An Enhanced Deep Learning Approach for Preventing Replay Attacks in Wireless Sensor Network

Abstract

The nature of Wireless Sensor Networks has established wide range of security threats due to minimal usage of hardware resources and nil infrastructures. Replay attack belongs to the category of Denial of service attack taking place in the network layer. The paper focuses on developing an enhanced deep learning approach for detecting and preventing the replay attack.  Decision trees along with assistance of support vector machine (SVM) is implemented on the dataset to prove its efficiency. Many research has proved that SVM has achieved above 98% success rate in detecting and preventing attacks on WSN. Solution for replay attack is less compared to other category of attacks belonging to Denial of service attack (DoS). A wireless sensor network is comprised of minimal hardware platform accelerated with radio communication involved in applications of agriculture and industry for measuring physical properties involved. In replay attack a piece of previously sent information is recorded and re-transmitted after some interval of time. The basic principle of this attack is adapted by more effective attacks like Sinkhole and Blackhole. The crucial role of this attack relies on incapacitating the effective functionality of the network. In this attack bogus packets covers the entire path from sensor node to base station. As a result of which a simulated time of propagation and fake signal strength is generated on receiver side where irrelevant location and fake distance is estimated based on arrival time of the signal. The best solution can be obtained by authentication accompanied with deep learning approach can be against replay attacks.