 |
Research Activities |
|
Research Activities |
My research interest falls in the areas of communication and security technologies of digital information. I have worked on many research problems in the related areas, however, my current research is focused on developing efficient algorithms for detecting anomalous events (anomalies) in an environment and experimental studies of this problem using distributed wireless sensor networks.
My previous research led to two novel algorithms and produced innovative benefits.
The first algorithm is a key generation and encryption algorithm for secure transmission of information across the Internet. It is a simple, flexible and computationally inexpensive algorithm that provides high security and scalability over a large number of Internet users. UNCG has filed a patent application in 2004 covering this algorithm. Australia, Singapore and Japan have awarded the patent for this innovative technology. UNCG and LiveCargo Inc., a Greensboro-based technology company, have signed an exclusive license and technical assistance agreement to develop and market this patent-pending security encryption technology. The second algorithm models the relationship between Internet-Router queue oscillation and Internet traffic characteristics. This algorithm reduces the chaotic queue oscillation at the Internet routers and provides better QoS to the users of Internet. It can be easily deployed in the current Internet-Routers for congestion control. Chaotic queue oscillation in Interne-Routers is an unsolved problem and the proposed algorithm is probably the first technique to significantly reduce chaotic queue oscillation problem in the commercial Interne-Routers.
Curently I am working on the following research Problem:
Anomaly detection is a vital technology for environment monitoring using Wireless Sensor Network (WSN) systems. With the anomaly detection technology, WSN systems can offer security and data integrity features by detecting anomalous events in the environment. The distinctive properties of genuine data and anomalies have been recently studied using scatter-plots of sensor data in vector-space and feature-space (kernel-based), and ellipsoid-based, multi-dimensional data approximation (MDDA) anomaly detection techniques have been developed. In vector-space the scatter-plots of raw sensor data of real-world applications do not display ellipsoidal patterns and hence it is difficult for ellipsoid-based MDDA techniques to construct robust ellipsoidal boundaries. In feature-space this problem is reduced but still nonlinear fuzzy, ellipsoidal boundaries exist with additional drawback of computational cost for constructing mathematical models. In our current project, we take this research into a new direction by defining a new form of feature-space. Our preliminary research with the new-space shows interesting ellipsoidal (Gaussian) patterns, data orientations and anomaly labeling capability. We investigate the properties of the scatter-plots of sensor data in multi-dimensional space and derive semantics for genuine data and anomalies for automatic labeling. We are also developing a robust, ellipsoid-based MDDA anomaly detection technique.
This research is important because it will fulfill the urgent need of monitoring abnormalities in many environment monitoring applications including health monitoring system, irrigation system and battlefield activity. The anomaly and genuine labeled data collected are disseminated to a wider community via the following websites http://www.issnip.unimelb.edu.au/research_program/downloads and http://www.uncg.edu/cmp/downloads.
