Security and privacy in resource constrained wireless networks

Wireless networks use radio waves as a communication medium which allows for faster and cheaper deployment. The networks being wireless, are out in the open, which makes them vulnerable to malicious users that can hinder their performance. Of the several types of wireless networks, we focus on security and privacy in wireless sensor networks (WSN) and cognitive radio mobile ad-hoc networks (CR-MANET). The devices in these networks are limited in resources such as energy, low power radio, etc. CR-MANET devices are mobile, requiring them to run on limited amount of energy supplied by batteries, and conserve energy by reducing communication cost using low power radios. In addition, sensor devices have limited storage and a slower CPU. The purpose of a WSN is to sense and report event occurrences, whereas a CR-MANET provides improved spectrum utilization. We studied three kinds of attacks on WSN. The first type of attack is on the source privacy of sensor nodes. This attack happens because an important characteristic of events detected by sensor devices is bound to the location of event occurrence that can be revealed by compromise of detecting sensor device's source privacy. Thus, protecting privacy of event detecting sensor device is of paramount importance for which we present an encryption based solution to protect source privacy under eavesdropping and node compromise attacks. The second type of attack by the malicious entity can be invasive in nature, which could possibly cause damage to the device, or can be passive as in side channel attacks. A comprehensive study of side channel attacks on WSN is presented, along with a process obfuscation technique. The third type of attack is on the propagation of data packet generated by the sensor device. The detected event data is sent to the base station. If a malicious entity is able to prevent such event reporting packets from reaching the base station and segregate the attack zone, it will be able to carry out its malicious activity without getting caught. To cover such scenarios, a proactive dynamic camouflage event generation solution is presented. CR-MANET devices sense for vacant licensed spectrum and improve its utilization in an opportunistic manner. Accurate licensed spectrum occupancy detection by a CR-MANET device is hampered by signal fading, hidden terminal problems, etc. Spectrum occupancy decision can be improved by cooperative spectrum sensing (CSS). However, CSS is made difficult by the presence of malicious users. The malicious users can have two goals: one is to disrupt the network, another is to manipulate the network for its own personal gains. The malicious users can create havoc in a CR-MANET by falsifying spectrum sensing information leading to interference with the primary user. The devices in a CR-MANET are mobile, which gives an opportunity for the malicious entity to hide behind the changing neighborhood. We present three solutions to overcome the spectrum sensing data falsification attack and incorrect reporting of signal measurement due to byzantine failures. The first is a multi-fusion based distributed spectrum sensing (MFDSS) using reputation propagation. In the second solution, a continuously evolving virtual neighbor cluster of past neighbor devices aid in validating the input gathered from the current neighboring devices (ReNVaS). Third, a recursive partitioning around medoids based clustering is performed to identify a tightly bound set of valid inputs for decision making (TMC). A unified and non-unified decision making strategy is presented using ReNVaS and TMC. MFDSS performs better in a fast changing network while performance of unified fusion is better in a slow mobility network with respect to primary user spectrum occupancy detection accuracy.