Entity

Time filter

Source Type


Suri N.N.R.R.,Center for and Robotics | Murty M.N.,Indian Institute of Science | Athithan G.,Center for and Robotics | Athithan G.,Scientific Analysis Group SAG
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2014

Many real world networks evolve over time indicating their dynamic nature to cope up with the changing real life scenarios. Detection of various categories of anomalies, also known as outliers, in graph representation of such network data is essential for discovering different irregular connectivity patterns with potential adverse effects such as intrusions into a computer network. Characterizing the behavior of such anomalies (outliers) during the evolution of the network over time is critical for their mitigation. In this context, a novel method for an effective characterization of network anomalies is proposed here by defining various categories of graph outliers depending on their temporal behavior noticeable across multiple instances of a network during its evolution. The efficacy of the proposed method is demonstrated through an experimental evaluation using various benchmark graph data sets. © 2014 Springer International Publishing. Source


Suri N.N.R.R.,Center for and Robotics | Murty M.N.,Indian Institute of Science | Athithan G.,Center for and Robotics | Athithan G.,Scientific Analysis Group SAG
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2013

Outlier detection is an important data mining task with applications in various domains. Mining of outliers in data has to deal with uncertainty regarding the membership of such outlier objects to one of the normal groups (classes) of objects. In this context, a soft computing approach based on rough sets happens to be a better choice to handle such mining tasks. Motivated by this requirement, a novel rough clustering algorithm is proposed here by modifying the basic k -modes algorithm to incorporate the lower and upper approximation properties of rough sets. The proposed algorithm includes the necessary computational steps required for determining the object assignment to various clusters and the modified centroid (mode) computation on categorical data. An experimental evaluation of the proposed rough k -modes algorithm is also presented here to demonstrate its performance in detecting outliers using various benchmark categorical data sets. © Springer-Verlag 2013. Source


Suri N.N.R.R.,Center for and Robotics | Murty M.N.,Indian Institute of Science | Athithan G.,Center for and Robotics | Athithan G.,Scientific Analysis Group SAG
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2013

Mining graph data has been an important data mining task due to its significance in network analysis and many other contemporary applications. Detecting anomalies in graph data is challenging due to the unsupervised nature of the problem and the size of the data itself to be dealt with. Recent research efforts in this direction have explored graph data for identifying anomalous nodes and anomalous edges of a given graph. However, in many real life applications where the data is inherently networked in nature, the requirement is to detect anomalous sub-graphs with distinguishing characteristics such as near cliques, etc. In this context, we propose a novel method for addressing the anomalous sub-graph mining problem through community detection by employing the non-negative matrix factorization technique. Anomalous sub-graphs are identified by applying some existing techniques on the detected communities for measuring their deviation from the normal characteristics. We demonstrate the effectiveness of the proposed method through experimental evaluation on various benchmark graph data sets. © Springer-Verlag 2013. Source


Mendiratta A.,Guru Tegh Bahadur Institute of Technology | Jha D.,Scientific Analysis Group SAG
International Conference on Electronics, Communication and Instrumentation 2014, ICECI 2014 | Year: 2014

The methods to controlling the noise in a signal have attracted many researchers over past few years. One such approach is Adaptive Noise Cancellation which has been proposed to reduce steady state additive noise. This method uses two inputs - primary and reference. The primary input receives signal from the signal source which has been corrupted with a noise uncorrelated to the signal. The reference input receives noise signal uncorrelated with the signal but correlated in some way to the noise signal in primary input. The reference input is adaptively filtered to obtain a close estimate of primary input noise which is then subtracted from the corrupted signal at the primary input to produce an estimate of a clean uncorrupted signal, which is the Adaptive Noise Cancellation output. A desired signal corrupted by noise can be recovered by feeding back this output to Adaptive Filter and implementing Least Mean Square algorithm to minimize output power. The audio signal corrupted with noise is used as a primary input and a noise signal is used as reference input. Computer simulations are carried out using MATLAB and experimental results are presented that illustrate the usefulness of Adaptive Noise Cancelling Technique. © 2014 IEEE. Source


Mishra M.,Banaras Hindu University | Chaturvedi U.,Banaras Hindu University | Pal S.K.,Scientific Analysis Group SAG
Souvenir of the 2014 IEEE International Advance Computing Conference, IACC 2014 | Year: 2014

Factorization of a number composed of two large prime numbers of almost equal number of digits is computationally a difficult task. The RSA public-key cryptosystem relies on this difficulty of factoring out the product of two very large prime numbers. There are various ways to find these two prime factors, but the huge memory and runtime expenses for large numbers pose tremendous difficulty. In this paper, we explore the possibility of solving this problem with the aid of Swarm Intelligence Metaheuristics using a Multithreaded Bound Varying Chaotic Firefly Algorithm. Firefly algorithm is one of the recent evolutionary computing models inspired by the behavior of fireflies. We have considered factors of equal number of digits. Observations show that the Firefly algorithm can be an effective tool to factorize a semi prime and hence can be further extended on extremely large numbers. © 2014 IEEE. Source

Discover hidden collaborations