A class-oriented feature selection approach for multi-class imbalanced network traffic datasets based on local and global metrics fusion

Liu, Zhen; Wang, Ruoyu; Tao, Ming; Cai, Xianfa

doi:10.1016/j.neucom.2015.05.089

Cited by 46 publications

(34 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The parameter w was decreased by 5% in each iteration while c 1 and c 2 was constant. According to the experiments described here, the PSO algorithm obtains the highest performance when w=1 and c c = = 2 1 2…”

Section: Particle Swarmmentioning

confidence: 89%

“…In the past few years, Support Vector Machines (SVMs) have shown excellent generalization power in classification problems in several application fields [1][2][3][4][5][6]. In addition to their strong theoretical background and high generalization ability, SVMs have been confirmed as a robust tool for classification and regression in several noisy and complex domains.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

PSO-based method for SVM classification on skewed data sets

et al. 2017

View full text Add to dashboard Cite

Over the last years, Support Vector Machines (SVMs) have become a successful approach in classification problems. However, the performance of SVMs is affected harshly by skewed data sets. An SVM learns a biased model that affects the performance of the classifier. Furthermore, SVMs are typically unsuccessful on data sets where the imbalanced ratio is very large. Lately, several techniques have been used to tackle this disadvantage by generating artificial instances. Artificial data instances attempt to add information to the minority class. However, the new instances could introduce noise and decrease the performance of the classifier. In this research, an alternative procedure is suggested, the algorithm finds systematically new instances, improving the performance of SVMs on skewed data sets. The proposed method starts obtaining the support vectors (SVs) from the skewed data set. These initial SVs are used to generate new instances and the PSO algorithm is used to evolve the artificial instances, eliminating noise instances. This research combines the best of optimization and classification techniques. To show the ability of the proposed method to improve the performance of SVMs on skewed data sets, we compare the performance of our method against some classical methods and show that our algorithm outperforms all of them on several data sets.

show abstract

Section: Particle Swarmmentioning

confidence: 89%

Section: Introductionmentioning

confidence: 99%

PSO-based method for SVM classification on skewed data sets

et al. 2017

View full text Add to dashboard Cite

show abstract

“…Then 30 tra±c values are created based on top 30 tra±c distribution at previous time bin and Eq. (11). Finally the created 30 tra±cs together with current top 30 tra±cs are sorted in descending order to determine the new top 30 tra±c.…”

Section: The Methods To Simulate Attack Tra±cmentioning

confidence: 99%

“…Besides, these methods need to use a large number of tra±c features to describe network tra±cs. At present, the anomaly detection and identi¯cation method based on tra±c feature selection 3,5,8,10,11,17,18 is widely used in the detection and classi¯cation of network attacks. The key of this kind of method is how to select appropriate numbers of tra±c features to describe di®erent types of network attacks.…”

Section: Introductionmentioning

confidence: 99%

Combining Renyi Entropy and EWMA to Detect Common Attacks in Network

Yan

2016

Int. J. Patt. Recogn. Artif. Intell.

View full text Add to dashboard Cite

How to timely and precisely identify attack behaviors in network without dealing with a large number of tra±c features and historical data, such as training data, is an important research work in the¯eld of network security. In this paper,¯rstly, the di®erences between Renyi entropy and Shannon entropy are analyzed and compared. In order to capture network tra±c changes exactly, Renyi entropy instead of Shannon entropy is proposed to measure selected tra±c features. Then EWMA control chart theory is used to check Renyi entropy time series for detecting and screening anomalies. And three kinds of network attacks are also analyzed and characterized by behavior feature vector for attack identi¯cation. Finally a feature similaritybased method is used to identify attacks. The experimental results of real tra±c traces show that the proposed method has good capability to detect and identify these attacks with less computation cost. To evaluate attack identi¯cation method conveniently, an approach is proposed to generate simulated attack tra±cs. Compared with Shannon entropy-based method, the experiments on simulation tra±cs show that Renyi entropy-based method has much higher overall accuracy, average precision and average true positive rate. Further comparison indicates the proposed method has more powerful performance to detect attacks than PCA-based method.

show abstract

“…The proposed method is applied on 22 less imbalanced and 22 high imbalanced datasets. Class Oriented Feature Selection (COFS) is proposed to improve the recall and precision for minority classes maintaining overall high accuracy in [24]. Cost sensitive learning is studied for multiclass imbalance problem in [25] and misclassification cost with respect to dependent class is investigated.…”

Section: Related Workmentioning

confidence: 99%

Multilayer Feedforward Neural Network for Internet Traffic Classification

Manju¹,

Harish²,

Nagadarshan³

2020

IJIMAI

View full text Add to dashboard Cite

Recently, the efficient internet traffic classification has gained attention in order to improve service quality in IP networks. But the problem with the existing solutions is to handle the imbalanced dataset which has high uneven distribution of flows between the classes. In this paper, we propose a multilayer feedforward neural network architecture to handle the high imbalanced dataset. In the proposed model, we used a variation of multilayer perceptron with 4 hidden layers (called as mountain mirror networks) which does the feature transformation effectively. To check the efficacy of the proposed model, we used Cambridge dataset which consists of 248 features spread across 10 classes. Experimentation is carried out for two variants of the same dataset which is a standard one and a derived subset. The proposed model achieved an accuracy of 99.08% for highly imbalanced dataset (standard).

show abstract

A class-oriented feature selection approach for multi-class imbalanced network traffic datasets based on local and global metrics fusion

Cited by 46 publications

References 38 publications

PSO-based method for SVM classification on skewed data sets

PSO-based method for SVM classification on skewed data sets

Combining Renyi Entropy and EWMA to Detect Common Attacks in Network

Multilayer Feedforward Neural Network for Internet Traffic Classification

Contact Info

Product

Resources

About