Narrow Results

The large amount of network traffic generated by Internet applications brings great challenges to Internet security. In order to facilitate network management and realize automatic classification of network traffic, this paper proposes a network traffic classification model NTCNET based on CNNs. Use open data set to do simulation verification experiment, then compare the test results with a variety of traditional classification methods. The experimental results shows that the constructed traffic classification model NTCNET has better precision, robustness and accuracy, with an accuracy of 99.66%.

This paper posits six new factors that impact customers’ expectations for next generation internet banking. The factors are innovative interface, integration with social media, money management tools, instant customer service, enhanced convenience, and next generation security. The data was collected from 310 respondents, who currently use internet banking services from leading countries. The results suggest that three factors, namely, integration with social media, money management tools, and enhanced convenience, would significantly affect customers’ expectations for next generation internet banking. The results also suggest the role of innovative interface, next generation security, and instant customer support and service in influencing the customers’ expectations. The proposed model would be useful for banks and technology providers to not only understand bank customers’ expectations, but also to launch innovative marketing strategies to gain competitive advantage, and to craft a unique selling proposition to its customers.

This article is motivated by internet security applications of multiple classifiers designed for the detection of malware. Following a standard approach in data mining, Dazeley et al. (Asian-European J. Math. 2 (2009)(1) 41–56) used Gröbner-Shirshov bases to define a family of multiple classifiers and develop an algorithm optimizing their properties.The present article complements and strengthens these results. We consider a broader construction of classifiers and develop a new and more general algorithm for the optimization of their essential properties.