José Everardo Bessa Maia

Network traffic prediction using PCA and K-means

The growing demand for link bandwidth and node capacity is a frequent phenomenon in IP network backbones. Within this context, traffic prediction is essential for the network operator. Traffic prediction can be undertaken based on link traffic or on origin-destination (OD) traffic which presents better results. This work investigates a methodology for traffic prediction based on multidimensional OD traffic, focusing on the stage of short-term traffic prediction using Principal Components Analysis as a technique for dimensionality reduction and a Local Linear Model based on K-means as a technique for prediction and trend analysis. The results validated with data on a real network present a satisfactory margin of error for use in practical situations.

Internet traffic classification using a Hidden Markov Model

This paper examines the performance of a new Hidden Markov Model (HMM) structure used as the core of an Internet traffic classsifier and compares the results against other models present in the literature. Traffic modeling and classification find importance in many areas such as bandwidth management, traffic analysis, prediction and engineering, network planning, Quality of Service provisioning and anomalous traffic detection. The new HMM structure, which takes into account the packet payload size (PS) and the inter-packet times (IPT) sequences, is obtained by concatenation of a first part which is framed with a HMM profile with another part whose structure is that of a fully-connected HMM. The first part captures the specific properties of the initial protocol packets while the second part captures the statistical properties of the whole sequence present in the flow. Models generated are found to increase the accurate in classifying different traffic classes in the analysed dataset. The average accuracy obtained by the classifier is 62.5% having seen only five packets, 80.0% after examining 13 packets and 95.5% after seeing the unidirectional entire flow.

Visual object tracking by an evolutionary self-organizing neural network

In this paper, a recently proposed evolutionary self-organizing map is extended and applied to visual tracking of objects in video sequences. The proposed approach uses a simple geometric template to track an object executing a smooth movement represented by affine transformations. The template is selected manually in the first frame and consists of a small number of keypoints and the neighborhood relations among them. The coordinates of the keypoints are used as the coordinates of the nodes of a non-regular grid defining a self-organizing map that represents the object. The weight vectors of each node in the output grid are updated by an evolutionary algorithm and used to locate the object frame by frame. Qualitative and quantitative evaluations indicate that the proposed approach present better results than those obtained by a direct method approach. Additionally, the proposed approach is evaluated under situations of partial occlusion and self-occlusion, and outliers, also presenting good results.

Online traffic classification based on sub-flows

Traffic classification by application class provides useful information for various tasks of network engineering and administration. However, offline classification of flows has limited its practical application to auditing tasks, long-term planning and other analytical issues. Therefore, research on traffic classification now moves towards the search for accurate and efficient methods of classification in order to meet online tasks such as traffic monitoring and shaping and other specific-application operations. In this work we apply the One-Against-All Approach (OAA) for two online classification strategies based on statistical features of TCP sub-flows. One uses the first N packets of the bi-directional TCP session and the other applies to sub-flows of the N packets starting at a random position in the flow. In our variant of the OAA approach, the problem of classifying an object in one of M classes is reduced to M binary classification problems with an associated decision rule, with each of them possibly using a different subset of features and sub-flow size. We investigated the effect of variation in the amount of N on the results of classification and the smaller set of variables in each of the above problems. This study used the Naïve Bayes classifier.

Improving Multidimensional Wireless Sensor Network Lifetime Using Pearson Correlation and Fractal Clustering

An efficient strategy for reducing message transmission in a wireless sensor network (WSN) is to group sensors by means of an abstraction denoted cluster. The key idea behind the cluster formation process is to identify a set of sensors whose sensed values present some data correlation. Nowadays, sensors are able to simultaneously sense multiple different physical phenomena, yielding in this way multidimensional data. This paper presents three methods for clustering sensors in WSNs whose sensors collect multidimensional data. The proposed approaches implement the concept of multidimensional behavioral clustering. To show the benefits introduced by the proposed methods, a prototype has been implemented and experiments have been carried out on real data. The results prove that the proposed methods decrease the amount of data flowing in the network and present low root-mean-square error (RMSE).

Fractal Clustering and similarity measure: Two new approaches for reducing energy consumption in Wireless Sensor Networks

Sensor clustering is an efficient strategy to reduce the number of messages flowing in a Wireless Sensor Network (WSN), decreasing, this way, the energy consumption in the network. This paper presents two new approaches for sensors clustering in WSNs, namely Fractal Clustering in Wireless Sensor Networks (FCWSN) and Similarity Measure in Wireless Sensor Networks (SMWSN). Both approaches are based on a new principle, known as behavioral clustering, which is able to cluster sensors with similar sensed data patterns of recent historical data collected. By exploring the new clustering method, the approaches are able to reduce message transmission by using cluster-heads for concentrating communication between sensors and sink. In order to validate and compare the proposed approaches, simulations have been conducted over real data, using SinalGo simulator. Results show that FCWSN and SMWSN can both significantly reduce the number of messages injected into the network whereas SMWSN presented a number of messages slightly smaller than FCWSN. In relation to Root Mean Square Error (RMSE), FCWSN remains about 10% lower than SMWSN approach, while both have a low RMSE.

Using fractal clustering to explore behavioral correlation: a new approach to reduce energy consumption in WSN

Sensor clustering is an efficient strategy to reduce the number of messages flowing in a Wireless Sensor Network (WSN) and thereby reducing the energy consumption. This paper presents a new approach to cluster sensors in WSNs, called Behavioral Correlation in WSN (BCWSN), which is based on the behavior of recent historical data collected by sensors. The proposed approach initializes clusters using the concepts of similarity in magnitude and trend of sensed data, and implements the notion of Fractal Clustering to dynamically find the best configuration for clusters. BCWSN can reduce the number of messages injected into the network when compared to approaches implementing temporal correlation, while RMSE remains roughly stable.

NTCS: A real time flow-based network traffic classification system

This work presents the design and implementation of a real time flow-based network traffic classification system. The classifier monitor acts as a pipeline consisting of three modules: packet capture and preprocessing, flow reassembly, and classification with Machine Learning (ML). The modules are built as concurrent processes with well defined data interfaces between them so that any module can be improved and updated independently. In this pipeline, the flow reassembly function becomes the bottleneck of the performance. In this implementation, was used a efficient method of reassembly which results in a average delivery delay of 0.49 seconds, aproximately. For the classification module, the performances of the K-Nearest Neighbor (KNN), C4.5 Decision Tree, Naive Bayes (NB), Flexible Naive Bayes (FNB) and AdaBoost Ensemble Learning Algorithm are compared in order to validate our approach.

A framework for processing complex queries in wireless sensor networks

In this work, we present a framework, denoted ADAGA -- P*, for processing complex queries and for managing sensor-field regression models. The proposed mechanism builds and instantiates sensor-field models. Thus ADAGA -- P* makes query engines able to answer complex queries such as give the probability of rain for the next two days in the city of Fortaleza. On the other hand, it is well known that minimizing energy consumption in a Wireless Sensor Network (WSN) is a critical issue for increasing the network lifetime. An efficient strategy for saving power in WSNs is to reduce the data volume injected into the network. For that reason, ADAGA -- P* implements an in-network data prediction mechanism in order to avoid that all sensed data have to be sent to fusion center node (or base station). Thus, sensor nodes only transmit data which are novelties for a regression model applied by ADAGA -- P*. Experiments using real data have been executed to validate our approach. The results show that ADAGA -- P* is quite efficient regarding communication cost and the number of executed float-point operations. In fact, the energy consumption rate to run ADAGA -- P* is up to 14 times lower than the energy consumed by kernel distributed regression for an RMSE difference of 0.003.

Evolving a self-organizing feature map for visual object tracking

An extension of a recently proposed evolutionary selforganizing map is introduced and applied to the tracking of objects in video sequences. In the proposed approach, a geometric template consisting of a small number of keypoints is used to track an object that moves smoothly. The coordinates of the keypoints and their neighborhood relations are associated with the coordinates of the nodes of a self-organizing map that represents the object. Parameters of a local affine transformation associated with each neuron are updated by an evolutionary algorithm and used to map each templates keypoint in the previous frame to the current one. Computer simulations indicate that the proposed approach presents better results than those obtained by a direct method approach.