Ihab Talkhan

Distributed Intrusion Detection System for Wireless Sensor Networks

Wireless Sensor Networks is extensively used in many of applications related to different fields. Some of those applications deal with confidential and critical data that must be protected from unauthorized access. Some other systems use WSNs that are deployed in very harsh environments with limited energy resources. Those systems cannot tolerate network failures that can be caused by network intruders. In this paper, an efficient intrusion detection model is introduced. The model uses intelligent techniques to detect intrusions. Two different architectures are introduced. The first architecture represents the level of sensor node, sink node, and base station. The second architecture represents the levels of sensor and sink nodes. This work proposes two intrusion detection algorithms, one uses a supervised learning mechanism to be used on the level of the sensor node and the other uses an unsupervised learning mechanism to be used on the levels of both the sink node and base station. The output of the algorithms is a set of detection rules which are structured in the form of binary tree. The introduced algorithms provided a high detection accuracy using less number of selected features, compared to previous work for intrusion detection, which decreases the complexity and the processing time. The proposed learning algorithms used only 10% of the data for training. An enhancement for J48 classification algorithm is also introduced which decreases the size of the algorithms decision tree and makes it suitable to be used for intrusion detection in WSNs.

Towards distributed layered intrusion detection system for large scale wireless sensor networks

In the current era of computer and communication rapid development, network security has become one of the most important factors to consider. Security considerations in wireless sensor networks (WSNs) have been an interesting point in research especially with the fast spread of WSNs. In this paper, an efficient two-layer and three-layer intrusion detection models are introduced. The two-layer model represents the levels of the sensor and sink nodes. The three-layer model represents the levels of the sensor, sink and base station. The models are elaborated and examined through a set of experiments. A supervised learning algorithm is introduced to be used in the sensor node layer and an unsupervised learning algorithm is introduced to be used in the other layers. The learning algorithms used only 10% of the data for training and gave a high detection accuracy on the used dataset, using lesser number of features compared to other approaches.

Residual Energy Considerations for AODV Enhancement

One of the main challenges in MANET is link breaks between the routes due to exhausted energy of nodes which have limited battery power. Every time a node transmits, receives or listens to a communication medium, it consumes energy. This paper focuses on creating an energy efficient method and a new technique was introduced to avoid link breaks before occurrence using preemptive local repair. The proposed technique measures residual energy of the node using two thresholds to determine the mode which the node uses to operate through, and hence provides prior information of link failure before its occurrence. Our proposed method is evaluated according to pause time and number of nodes in network. The results showed that our method improved the performance of AODV as the number of delivered packets is increased, the number of broken links is decreased and the total energy of the network is increased.

Corner-based background segmentation using Adaptive Resonance Theory

A correct video segmentation, namely the detection of moving objects within a scene plays a very important role in many application in safety, surveillance, traffic monitoring and object detection. The main objective of this paper is to implement an effective background segmentation algorithm for corner sets extracted from video sequences. A dynamic prototype of the structure of background corners is produced and incoming corners are classified using a Fuzzy ARTMAP Neural Network and labeled as pertaining to the background or foreground using a spatial clustering method. Finally the accuracy of the proposed algorithm is evaluated using PETS2006 benchmark data.

Comparing the Performance of Learnable Evolution Model LEM and Pattern Search as a Function Optimizer

The underlying paper presents a comparison of the learnable evolution model LEM and Pattern Search PS techniques as a function optimizer. In contrast to conventional Darwinian type evolutionary computation algorithm that uses various forms of mutation and/or recombination operators, LEM uses machine learning to guide the process of generating new individuals. It employs the AQ learning to generate hypotheses discriminating between groups of high and low fitness individuals, and then uses these hypotheses to generate new individuals. On the other hand pattern search is a class of direct search for derivative-free optimization with accurately established global convergence properties. Pattern search makes no use of derivative information, which might be unavailable, too expensive, or misleading. This paper focuses on measuring the performance of LEM3 and pattern search from the point of view of execution time in experiments on optimizing the Rastrigin function with different number of variables.