Radhakrishnan Maivizhi
@nitpy.ac.in
Faculty (On Contract), CSE
National Institute of Technology Puducherry, Karaikal, Puducherry, India
RESEARCH INTERESTS
Wireless Sensor Networks
Internet of Things
Machine Learning
Network Security
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Radhakrishnan Maivizhi and Palanichamy Yogesh
Inderscience Publishers
Radhakrishnan Maivizhi and Palanichamy Yogesh
Springer Science and Business Media LLC
Radhakrishnan Maivizhi and Palanichamy Yogesh
Springer Science and Business Media LLC
Radhakrishnan Maivizhi and Palanichamy Yogesh
IEEE
The dense distributed deployment of Wireless Sensor Networks (WSNs) causes the sensors to generate big amount of data which are highly correlated and redundant. Transmitting such spatially correlated and redundant data consumes more sensor energy and consequently decreases the lifetime of network. Eliminating redundancy is thus necessary in the sensed data and also while processing the sensed data. By employing appropriate data aggregation techniques, the data redundancy can be minimized. We leverage statistical techniques in sensor networks and propose a novel Spatial Correlation based Data Redundancy Elimination for Data Aggregation (SCDRE) protocol that eliminates redundancy at two levels: at source level, it uses simple data similarity function and at aggregator level, it uses correlation coefficient to eliminate redundancy and aggregate the data. We have evaluated the proposed protocol in terms of aggregation ratio, data accuracy and energy consumption and the results show that SCDRE outperforms other existing techniques. In addition, SCDRE is more robust in the presence of noise and outliers. By eliminating data redundancy to greater extent, our protocol experiences less communication overhead and significantly enhances the lifetime of sensor networks.
Radhakrishnan Maivizhi and Palanichamy Yogesh
ACM
Wireless sensor networks (WSNs) deployed in a plethora of applications produce a significant portion of big data. Handling these huge volume of data is a critical challenge in a resource constrained wireless sensor networks. Data aggregation is the most practical and important paradigm in big data wireless sensor networks. It reduces the huge volume of data by combining the similar data and eliminating data redundancy and reduces thereby the resource consumption. However preserving data confidentiality and integrity along with en-route aggregation is a great challenge. In this paper, we propose a novel Concealed Multidimensional Data Aggregation (CMDA) protocol for big data wireless sensor networks. CMDA integrates super-increasing sequence and homomorphic encryption to structure the multidimensional data and protect the data privacy and a homomorphic signature to check the integrity of data. In addition, the proposed protocol filters false data packets and achieves data freshness. Security analysis reveals that the proposed protocol achieves end-to-end security and performance evaluation shows that CMDA incurs less communication overhead and consequently reduces energy consumption which enhances the lifetime of sensor networks. To the best of our knowledge, this is the first work that achieves end-to-end security in multidimensional data aggregation.
Radhakrishnan Maivizhi and Palanichamy Yogesh
ACM
1 EXTENDED ABSTRACT Data aggregation or in-network aggregation plays a fundamental and vital role in maximizing the lifetime of wireless sensor networks (WSNs). However, in big data applications, existing data aggregation (DA) techniques based on compressed sensing, discrete cosine tranform and principal component analysis [2] suffer from problems such as high energy consumption and complex data analysis. To overcome these problems, we propose a novel linear discriminant analysis based data aggregation protocol for multidimensional data in big data WSNs. The proposed protocol employs Fisher Linear Discriminant (FLD) [1], a machine learning technique for aggregating multidimensional data. Researchers defined FLD as a classifier and a dimensionality reduction technique. Classifier: Data aggregation is performed after projecting the multidimensional data down to one dimension as follows. y =WT x ≥ −w0 (1)
Radhakrishnan Maivizhi and Palanichamy Yogesh
IGI Global
In-network aggregation is a natural approach in wireless sensor networks (WSNs) to collaboratively process data generated by the sensor nodes. Besides processing, in-network aggregation also achieves effective energy consumption and bandwidth utilization. Since the sensing devices of a WSN are prone to a variety of attacks due to wireless communication and limited resources, secure in-network aggregation is a great challenge. This article proposes a secure in-network aggregation (SINA) protocol for additive aggregation functions. This protocol integrates privacy homomorphism (PH) and secret sharing to achieve both data confidentiality and data integrity. Additionally, the proposed protocol ensures message authentication and data freshness. Moreover, it achieves false data screening in-network should be changed as in-network false data screening which considerably saves energy by not transmitting false packets. Security analysis reveals that SINA protects the network from variety of attacks. Performance analysis shows that SINA consumes less energy while achieving end-to-end security, and thereby increases the lifetime of the WSN.
Maivizhi Radhakrishnan and Yogesh Palanichamy
IEEE
The primary objective of Concealed Data Aggregation (CDA) is to preserve privacy while performing enroute data aggregation. Privacy Homomorphism (PH) techniques facilitate the development of concealed data aggregation protocols that enable end-to-end encryption in wireless sensor networks. Nevertheless, these protocols do not satisfy other security requirements such as integrity, authentication and data freshness and are vulnerable to various attacks. In this paper, we propose the novel Intrusion Resilient Concealed Data Aggregation (IRCDA) protocol for secure in-network data aggregation. This protocol employs privacy homomorphism techniques to achieve data privacy and combines them with homomorphic digital signatures to protect the integrity of the aggregated result. In addition, the proposed approach is resilient against node compromise attacks, an attack that devastates the function of Wireless Sensor Networks (WSNs). Security analysis and performance evaluation show that ours is the first work which remains secure even after compromising many sensor nodes and base station though not simultaneously. It completely mitigates the harmful effect caused by key exposure. Moreover, our protocol incurs very less communication overhead with slight increase in computation which is negligible when compared to communication cost and thereby it prolongs the lifetime of the network. The results prove that IRCDA performs better than most of the existing techniques in providing security to the process of data aggregation.
Radhakrishnan Maivizhi and Palanichamy Yogesh
Springer Singapore
R. Maivizhi, S. Sendhilkumar, and G. S. Mahalakshmi
ACM
Online Social Networks (OSNs) namely Facebook, Twitter and LinkedIn are the most popularly visited sites on the internet. These sites contain large voluminous data about the people and relationships among them. Community structure is an important property of social networks. It is a topic of considerable interest in many areas due to its wide range of applications in multiple disciplines including biology, computer science, social sciences and so on. Detection of communities reveals how the structure of ties affects the peoples and their relationships. To facilitate community discovery a wide range of tools have been developed over years. This paper surveys several tools available for detection and mining of communities and presents a comparative study. In addition, we discussed various visualization layouts of social networks in order to perceive network data and to communicate the result of analysis.
Maivizhi R. and Matilda S.
IEEE
Wireless networks are prone to identity based spoofing attacks and tend to degrade network performance. Received Signal Strength (RSS) spatial correlation is usually used for detecting and localizing spoofing attacks. This cannot be applied to environments where RSS value is not stable and varies with distance. This approach is also not desirable for accurate localization of multiple adversaries. This paper proposes Distance based Detection and Localization (DDL) algorithm that adds distance parameter to the existing system to perform spoofing detection and accurate localization of multiple adversaries. In addition, it determines the number of attackers, eliminates them from the network and thereby improves network performance. Simulation results demonstrate that this proposed work provides excellent localization performance and is generic across different technologies including IEEE 802.11 (WiFi) and IEEE 802.15.4 (ZigBee) networks.