Wireless Sensor Network Security: Real-Time Detection and Prevention of Attacks

Security is a critical issue in many real-world applications of wireless sensor networks (WSNs). This research focuses on implementing security mechanisms, against two specific types of attack that occur in a network of MICAz motes. These are ´denial-of-service´ (DoS) attacks where unnecessary packets are sent causing services to appear unavailable and thus these services are denied to the legitimate sensor nodes, and ´passive information gathering´ attacks where an adversary tries to obtain the confidential information stored in wireless sensor nodes. A machine learning technique named neural network (NN) is used to detect DoS attack conducted by an adversary. We compare and evaluate NN with our new method. The results clearly show an improved performance for our new proposed symmetric-key algorithm compared to NN. Lower encryption computational energy cost (8 μJ vs 16 μJ), lower memory requirements (8896 bytes of ROM and 434 bytes of RAM vs 15848 bytes of ROM and 763 bytes of RAM), and less execution time (0.164835 ms vs 0.6208791 ms) show the significant advantages of using symmetric-key algorithm instead of NN for detecting DoS attack.