Mubasher Baig - Academia.edu (original) (raw)

Papers by Mubasher Baig

IEEE Access

Adaptive Boosting (AdaBoost) based meta learning algorithms generate an accurate classifier ensem... more Adaptive Boosting (AdaBoost) based meta learning algorithms generate an accurate classifier ensemble using a learning algorithm with only moderate accuracy guarantees. These algorithms have been designed to work in typical supervised learning settings and hence use only labeled training data along with a base learning algorithm to form an ensemble. However, significant knowledge about the solution space might be available along with training data. The accuracy and convergence rate of AdaBoost might be improved using such knowledge. An effective way to incorporate such knowledge into boosting based ensemble learning algorithms is presented in this paper. Using several synthetic and real datasets, empirical evidence is reported to show the effectiveness of proposed method.Significant improvements have been obtained by applying the proposed method for detecting roads in aerial images.

Determining the state of each cell, for instance, cell outages, in a densely deployed cellular ne... more Determining the state of each cell, for instance, cell outages, in a densely deployed cellular network is a difficult problem. Several prior studies have used minimization of drive test (MDT) reports to detect cell outages. In this paper, we propose a two step process. First, using the MDT reports, we estimate the serving base station’s transmit power for each user. Second, we learn summary statistics of estimated transmit power for various networks states and use these to classify the network state on test data. Our approach is able to achieve an accuracy of 96% on an NS-3 simulation dataset. Decision tree, random forest and SVM classifiers were able to achieve a classification accuracy of 72.3%, 76.52% and 77.48%, respectively .

Journal of Intelligent & Fuzzy Systems, 2017

This paper presents a cascade of ensemble-based artificial neural network for multi-class intrusi... more This paper presents a cascade of ensemble-based artificial neural network for multi-class intrusion detection (CANID) in computer network traffic. The proposed system learns a number of neural-networks connected as a cascade with each network trained using a small sample of training examples. The proposed cascade structure uses the trained neural network as a filter to partition the training data and hence a relatively small sample of training examples are used along with a boosting-based learning algorithm to learn an optimal set of neural network parameters for each successive partition. The performance of the proposed approach is evaluated and compared on the standard KDD CUP 1999 dataset as well as a very recent dataset, UNSW-NB15, composed of contemporary synthesized attack activities. Experimental results show that our proposed approach can efficiently detect various types of cyber attacks in computer networks.

Lecture Notes in Computer Science, 2015

A novel method for learning a linear multilayer feedforward artificial neural network (ANN) by us... more A novel method for learning a linear multilayer feedforward artificial neural network (ANN) by using ensembles of boosted decision stumps is presented. Network parameters are adapted through a layer-wise iterative traversal of neurons with weights of each neuron learned by using a boosting based ensemble and an appropriate reduction. Performances of several neural network models using the proposed method are compared for a variety of datasets with networks learned using three other algorithms, namely Perceptron learning rule, gradient decent back propagation algorithm, and Boostron learning.

Lecture Notes in Computer Science, 2014

ABSTRACT A novel boosting based perceptron learning algorithm is presented that uses AdaBoost alo... more ABSTRACT A novel boosting based perceptron learning algorithm is presented that uses AdaBoost along with a new representation of decision stumps using homogenous coordinates. The new representation of decision stumps makes perceptron an instance of boosting based ensemble. As Boostron minimizes an exponential cost function instead of the mean squared error minimized by the perceptron learning algorithm, it gives improved performance for classification problems. The proposed method is compared to the perceptron learning algorithm using several classification problems of varying complexity.

2014 International Joint Conference on Neural Networks (IJCNN), 2014

A boosting-based cascade for automatic decomposition of multiclass learning problems into several... more A boosting-based cascade for automatic decomposition of multiclass learning problems into several binary classification problems is presented. The proposed cascade structure uses a boosted classifier at each level and use a filtering process to reduce the problem size at each level. The method has been used for detecting malicious traffic patterns using a benchmark intrusion detection dataset. A comparison of the approach with four boosting-based multiclass learning algorithms is also provided on this dataset.

Lecture Notes in Computer Science, 2012

ABSTRACT Boosting is a generic statistical process for generating accurate classifier ensembles f... more ABSTRACT Boosting is a generic statistical process for generating accurate classifier ensembles from moderately accurate learning algorithm. This paper presents a new generic boosting style procedure, M-Boost , for learning multiclass concepts. M-Boost uses a global strategy for selecting the weak classifier, a global weight reassignment strategy, a vector valued weight for the selected classifiers, and an ensemble that outputs a probability distribution on classes.

IEICE Transactions on Communications, 2008

In this letter, we propose effective feature vectors to improve the performance of voice activity... more In this letter, we propose effective feature vectors to improve the performance of voice activity detection (VAD) employing a support vector machine (SVM), which is known to incorporate an optimized nonlinear decision over two different classes. To extract the effective feature vectors, we present a novel scheme that combines the a posteriori SNR, a priori SNR, and predicted SNR, widely adopted in conventional statistical model-based VAD.

Cryptography is an exciting field of knowledge that deals with secure transmission of data on pub... more Cryptography is an exciting field of knowledge that deals with secure transmission of data on public channels. It has found many applications in the digital era and has attracted some of the greatest minds of the world. Many good books on cryptography and cryptographic protocol are available and more are being written. The field of information theory is also very important and highly applicable area. As pointed out by Thomas M. Cover and Joy A. Thomas in their famous book, Elements of Information Theory, that information theory is all about computing ultimate compression and computing maximum possible data rates on a communication channel. The theory of error-correcting codes deals with the error free transmission of messages on noisy channels. Therefore, all the three fields, obviously, deal with transmission of messages and are interrelated. Cryptography, Information Theory and Error-Correction is an excellent introduction to the three interrelated fields. As pointed out in its preface, the book is intended to provide a complete but highly accessible account of the three subjects and their interconnection. The presentation of the material by Bruen and Forcinito is very attractive and easy to follow. It is infact highly accessible even for an undergraduate student. The book contains 24 chapters and is naturally divided into three parts, one each for the three main topics. The first eight chapters discuss the basics of cryptography and constitute the first part of the book. The next nine chapters are devoted to the discussion of information theory and its applications and the last 7 chapters contain the theory of error correcting codes and the algorithm proposed by the authors. Part I of the book starts with a brief introduction of cryptography and the biography of Claude E. Shannon, and it then goes on describing the classical enciphering techniques (like Caesar and Vigenere ciphers), symmetric key and public key cryptosystems (like RSA, DES and AES), different security protocols (like SSL, PGP and GPG), the ideas of digital signatures, hash functions, key exchange and key management systems like Kerberos. Elliptic curve cryptography and different cryptanalysis techniques are also included in this part. This section ends with a brief discussion of some practical issues, including technical, commercial and legal aspects in the development of practical systems. First 5 chapters in this part are very well written. The authors have demonstrated good presentation and teaching skills in these chapters. The description of RSA in chapter 3, DES rounds in chapter 4 and AES transformations in chapter 4 are interesting and easy to follow. Although a complete chapter is devoted for Elliptic Curve Cryptography (ECC) but the treatment of various uses of ECC is not as detailed. Given that most of the resent research in mathematics went into the study of elliptic curves (as described by the authors in the book) the material given in this chapter is very brief. Most of the chapter is devoted to describing the group structure provided by the elliptic curves and in introducing the details of arithmetic operations defined on an elliptic curve. Like ECC, a full chapter is devoted to the description of cryptanalysis techniques. Given that cryptanalysis is one of the most interesting parts of cryptography, the level of detail given in this chapter does not seem to be

IEEE Access

Adaptive Boosting (AdaBoost) based meta learning algorithms generate an accurate classifier ensem... more Adaptive Boosting (AdaBoost) based meta learning algorithms generate an accurate classifier ensemble using a learning algorithm with only moderate accuracy guarantees. These algorithms have been designed to work in typical supervised learning settings and hence use only labeled training data along with a base learning algorithm to form an ensemble. However, significant knowledge about the solution space might be available along with training data. The accuracy and convergence rate of AdaBoost might be improved using such knowledge. An effective way to incorporate such knowledge into boosting based ensemble learning algorithms is presented in this paper. Using several synthetic and real datasets, empirical evidence is reported to show the effectiveness of proposed method.Significant improvements have been obtained by applying the proposed method for detecting roads in aerial images.

Determining the state of each cell, for instance, cell outages, in a densely deployed cellular ne... more Determining the state of each cell, for instance, cell outages, in a densely deployed cellular network is a difficult problem. Several prior studies have used minimization of drive test (MDT) reports to detect cell outages. In this paper, we propose a two step process. First, using the MDT reports, we estimate the serving base station’s transmit power for each user. Second, we learn summary statistics of estimated transmit power for various networks states and use these to classify the network state on test data. Our approach is able to achieve an accuracy of 96% on an NS-3 simulation dataset. Decision tree, random forest and SVM classifiers were able to achieve a classification accuracy of 72.3%, 76.52% and 77.48%, respectively .

Journal of Intelligent & Fuzzy Systems, 2017

This paper presents a cascade of ensemble-based artificial neural network for multi-class intrusi... more This paper presents a cascade of ensemble-based artificial neural network for multi-class intrusion detection (CANID) in computer network traffic. The proposed system learns a number of neural-networks connected as a cascade with each network trained using a small sample of training examples. The proposed cascade structure uses the trained neural network as a filter to partition the training data and hence a relatively small sample of training examples are used along with a boosting-based learning algorithm to learn an optimal set of neural network parameters for each successive partition. The performance of the proposed approach is evaluated and compared on the standard KDD CUP 1999 dataset as well as a very recent dataset, UNSW-NB15, composed of contemporary synthesized attack activities. Experimental results show that our proposed approach can efficiently detect various types of cyber attacks in computer networks.

Lecture Notes in Computer Science, 2015

A novel method for learning a linear multilayer feedforward artificial neural network (ANN) by us... more A novel method for learning a linear multilayer feedforward artificial neural network (ANN) by using ensembles of boosted decision stumps is presented. Network parameters are adapted through a layer-wise iterative traversal of neurons with weights of each neuron learned by using a boosting based ensemble and an appropriate reduction. Performances of several neural network models using the proposed method are compared for a variety of datasets with networks learned using three other algorithms, namely Perceptron learning rule, gradient decent back propagation algorithm, and Boostron learning.

Lecture Notes in Computer Science, 2014

ABSTRACT A novel boosting based perceptron learning algorithm is presented that uses AdaBoost alo... more ABSTRACT A novel boosting based perceptron learning algorithm is presented that uses AdaBoost along with a new representation of decision stumps using homogenous coordinates. The new representation of decision stumps makes perceptron an instance of boosting based ensemble. As Boostron minimizes an exponential cost function instead of the mean squared error minimized by the perceptron learning algorithm, it gives improved performance for classification problems. The proposed method is compared to the perceptron learning algorithm using several classification problems of varying complexity.

2014 International Joint Conference on Neural Networks (IJCNN), 2014

A boosting-based cascade for automatic decomposition of multiclass learning problems into several... more A boosting-based cascade for automatic decomposition of multiclass learning problems into several binary classification problems is presented. The proposed cascade structure uses a boosted classifier at each level and use a filtering process to reduce the problem size at each level. The method has been used for detecting malicious traffic patterns using a benchmark intrusion detection dataset. A comparison of the approach with four boosting-based multiclass learning algorithms is also provided on this dataset.

Lecture Notes in Computer Science, 2012

ABSTRACT Boosting is a generic statistical process for generating accurate classifier ensembles f... more ABSTRACT Boosting is a generic statistical process for generating accurate classifier ensembles from moderately accurate learning algorithm. This paper presents a new generic boosting style procedure, M-Boost , for learning multiclass concepts. M-Boost uses a global strategy for selecting the weak classifier, a global weight reassignment strategy, a vector valued weight for the selected classifiers, and an ensemble that outputs a probability distribution on classes.

IEICE Transactions on Communications, 2008

In this letter, we propose effective feature vectors to improve the performance of voice activity... more In this letter, we propose effective feature vectors to improve the performance of voice activity detection (VAD) employing a support vector machine (SVM), which is known to incorporate an optimized nonlinear decision over two different classes. To extract the effective feature vectors, we present a novel scheme that combines the a posteriori SNR, a priori SNR, and predicted SNR, widely adopted in conventional statistical model-based VAD.

Cryptography is an exciting field of knowledge that deals with secure transmission of data on pub... more Cryptography is an exciting field of knowledge that deals with secure transmission of data on public channels. It has found many applications in the digital era and has attracted some of the greatest minds of the world. Many good books on cryptography and cryptographic protocol are available and more are being written. The field of information theory is also very important and highly applicable area. As pointed out by Thomas M. Cover and Joy A. Thomas in their famous book, Elements of Information Theory, that information theory is all about computing ultimate compression and computing maximum possible data rates on a communication channel. The theory of error-correcting codes deals with the error free transmission of messages on noisy channels. Therefore, all the three fields, obviously, deal with transmission of messages and are interrelated. Cryptography, Information Theory and Error-Correction is an excellent introduction to the three interrelated fields. As pointed out in its preface, the book is intended to provide a complete but highly accessible account of the three subjects and their interconnection. The presentation of the material by Bruen and Forcinito is very attractive and easy to follow. It is infact highly accessible even for an undergraduate student. The book contains 24 chapters and is naturally divided into three parts, one each for the three main topics. The first eight chapters discuss the basics of cryptography and constitute the first part of the book. The next nine chapters are devoted to the discussion of information theory and its applications and the last 7 chapters contain the theory of error correcting codes and the algorithm proposed by the authors. Part I of the book starts with a brief introduction of cryptography and the biography of Claude E. Shannon, and it then goes on describing the classical enciphering techniques (like Caesar and Vigenere ciphers), symmetric key and public key cryptosystems (like RSA, DES and AES), different security protocols (like SSL, PGP and GPG), the ideas of digital signatures, hash functions, key exchange and key management systems like Kerberos. Elliptic curve cryptography and different cryptanalysis techniques are also included in this part. This section ends with a brief discussion of some practical issues, including technical, commercial and legal aspects in the development of practical systems. First 5 chapters in this part are very well written. The authors have demonstrated good presentation and teaching skills in these chapters. The description of RSA in chapter 3, DES rounds in chapter 4 and AES transformations in chapter 4 are interesting and easy to follow. Although a complete chapter is devoted for Elliptic Curve Cryptography (ECC) but the treatment of various uses of ECC is not as detailed. Given that most of the resent research in mathematics went into the study of elliptic curves (as described by the authors in the book) the material given in this chapter is very brief. Most of the chapter is devoted to describing the group structure provided by the elliptic curves and in introducing the details of arithmetic operations defined on an elliptic curve. Like ECC, a full chapter is devoted to the description of cryptanalysis techniques. Given that cryptanalysis is one of the most interesting parts of cryptography, the level of detail given in this chapter does not seem to be