Cognitive security framework for heterogeneous sensor network using swarm intelligence
Abstract
Rapid development of sensor technology has led to applications ranging from academic to military in a short time span. These tiny sensors are deployed in environments where security for data or hardware cannot be guaranteed. Due to resource constraints, traditional security schemes cannot be directly applied. Unfortunately, due to minimal or no communication security schemes, the data, link and the sensor node can be easily tampered by intruder attacks.
This dissertation presents a security framework applied to a sensor network that can be managed by a cohesive sensor manager. A simple framework that can support security based on situation assessment is best suited for chaotic and harsh environments. The objective of this research is designing an evolutionary algorithm with controllable parameters to solve existing and new security threats in a heterogeneous communication network. An in-depth analysis of the different threats and the security measures applied considering the resource constrained network is explored. Any framework works best, if the correlated or orthogonal performance parameters are carefully considered based on system goals and functions. Hence, a trade-off between the different performance parameters based on weights from partially ordered sets is applied to satisfy application specific requirements and security measures.
The proposed novel framework controls heterogeneous sensor network requirements, and balance the resources optimally and efficiently while communicating securely using a multi-objection function. In addition, the framework can measure the affect of single or combined denial of service attacks and also predict new attacks under both cooperative and non-cooperative sensor nodes. The cognitive intuition of the framework is evaluated under different simulated real time scenarios such as Health-care monitoring, Emergency Responder, VANET, Biometric security access system, and Battlefield monitoring.
The proposed three-tiered Cognitive Security Framework is capable of performing situation assessment and performs the appropriate security measures to maintain reliability and security of the system. The first tier of the proposed framework, a cross-layer cognitive security protocol defends the communication link between nodes during denial-of-Service attacks by re-routing data through secure nodes. The cognitive nature of the protocol balances resources and security making optimal decisions to obtain reachable and reliable solutions. The versatility and robustness of the protocol is justified by the results obtained in simulating health-care and emergency responder applications under Sybil and Wormhole attacks. The protocol considers metrics from each layer of the network model to obtain an optimal and feasible resource efficient solution.
In the second tier, the emergent behavior of the protocol is further extended to mine information from the nodes to defend the network against denial-of-service attack using Bayesian models. The jammer attack is considered the most vulnerable attack, and therefore simulated vehicular ad-hoc network is experimented with varied types of jammer. Classification of the jammer under various attack scenarios is formulated to predict the genuineness of the attacks on the sensor nodes using receiver operating characteristics. In addition to detecting the jammer attack, a simple technique of locating the jammer under cooperative nodes is implemented. This feature enables the network in isolating the jammer or the reputation of node is affected, thus removing the malicious node from participating in future routes.
Finally, a intrusion detection system using 'bait' architecture is analyzed where resources is traded-off for the sake of security due to sensitivity of the application. The architecture strategically enables ant agents to detect and track the intruders threatening the network. The proposed framework is evaluated based on accuracy and speed of intrusion detection before the network is compromised. This process of detecting the intrusion earlier helps learn future attacks, but also serves as a defense countermeasure. The simulated scenarios of this dissertation show that Cognitive Security Framework is best suited for both homogeneous and heterogeneous sensor networks.
