Rozmie R. Othman

Network traffic classification — A comparative study of two common decision tree methods: C4.5 and Random forest

Network traffic classification gains continuous interesting while many applications emerge on the different kinds of networks with obfuscation techniques. Decision tree is a supervised machine learning method used widely to identify and classify network traffic. In this paper, we introduce a comparative study focusing on two common decision tree methods namely: C4.5 and Random forest. The study offers comparative results in two different factors are accuracy of classification and processing time. C4.5 achieved high percentage of classification accuracy reach to 99.67 for 24000 instances while Random Forest was faster than C4.5 in term of processing time.

Practical Adoptions of T-Way Strategies for Interaction Testing

This paper discusses the practical adoption of t-way strategies (also termed interaction testing) for interaction testing. Unlike earlier work, this paper also highlights and unifies the different possible use of t-way strategies including uniform interaction, variable strength interaction, and input-output based relations. In order to help engineers make informed decision on the different use of t-way strategies, this paper discusses the main issues and shortcomings to be considered as well as demonstrates some practical results with a-step-by-step example. In doing so, this paper also analyzes the related works highlighting the current state-of-the-arts and capabilities of some of the existing t-way strategy implementations.

Event Driven Input Sequence T-Way Test Strategy Using Simulated Annealing

Exhaustive testing for event driven sequence input interaction is costly and not always practicable for all types of software testing. So, an alternative technique is crucial where optimum/near optimum test case generation is key concern. This paper presents a feasible test suite generation technique using Simulated Annealing (SA) algorithm for Event Driven Input Sequence Testing (EDIST)and called EDIST-SA T-way test strategy. The EDIST-SA technique focuses on a heuristic analysis for generating feasible and near optimum test suite(s), where a cost function carefully initiates acceptable test candidates and a fine-tuned cooling rate with temperature takes part as an iterative perspective. We corroborate on EDIST-SA algorithm by doing a large number of experiments to achieve optimum and/or near optimum test cases from a large number of test candidates. The experimental results are tested on a real application called Embedded Network Traffic Monitoring System (ENTM). Analysis on EDIST-SA strategy shows that the optimum test suite is found from some of the iterated solution and there is possibility to have more feasible accepted test suites.

The effects of CPU load & idle state on embedded processor energy usage

Device power consumption is a serious design consideration especially for embedded systems. By reducing the power consumption of a particular system, we could effectively prolong the runtime of the system, allowing for longer operational condition of a particular system. Previous studies have suggested that the power characteristics of a modern embedded processor have since been improved with manufacturers implementation of better energy-focused designs. Implementation of hardware optimization such as better clock and power gating have been shown to produce better energy usage during on-load and off-load processing. In this paper we benchmarked the energy use of a modern embedded processor and study the effects of idling time to the processor and system energy usage. We have found that the processor energy use is significantly reduced in the instant that the processor goes idle during the execution process. The idling time during a processing timeslice allows the processor to use significantly less energy without explicitly depending on a frequency scaling algorithm to reduce energy consumption. This power saving feature directly implemented inside the processor hardware have the possibility to render software based frequency scaling algorithm and DVFS method to be less effective in reducing energy usage.

A Performance Study of Hidden Markov Model and Random Forest in Internet Traffic Classification

Currently, the emphasis on internet traffic classification is gaining momentum and a broad range of procedures has been developed for a variety of network tools. Machine learning methods are commonly utilized for the recognition and classification of internet traffic. This investigation trains its sights on two frequently employed techniques in this area. These are the Hidden Markov Model and the Random Forest procedure. Furthermore, the study offers a detailed explanation and analysis for the methodologies for each and forwards the advantages and setbacks attributed through comparing the performance in different significant factors.

Analysis of Geometric, Zernike and United Moment Invariants Techniques Based on Intra-class Evaluation

In this paper, three moment invariants techniques have been used to extract the shape properties of the image. There are geometric moment, zernike moment and united moment invariants. These moment invariants have been used to analyze the image due to its invariant features of an image based on scaling factor and rotation. A set of equations known as intra-class analysis has been applied to measure the similarity of feature vector that represent the same object. The results obtained in this study have been analyzed and compared in terms of intra-class analysis in order to find the best technique among the three different types of moments. Based on the results that have been obtained by using the similar image, it is found that the geometric and united moment invariants techniques are better with small values of total percentage mean absolute error (TPMAE) as compared to zernike moment invariants.

Input-input relationship constraints in T-way testing

T-way testing is designed to detect faults due to interaction. In order to be effective, all t combinations of input parameters must be tested. While many t-way strategies can be used to generate the t-way test data (e.g. IPOG, AETG, GT-Way, Jenny, TVG and MIPOG), most do not ensure that all t combinations of input parameters can be practically tested. Addressing this issue, this paper highlights a new type of constraints that might prevent some t-way parameter interactions from being tested (and hence compromising the effectiveness of t-way testing), termed input-input relationship constraints. Apart from ensuring all t combinations are properly tested, input-input relationship constraints can further optimize the generated test data since all impossible combinations are completely ignored. In addition, this paper also introduces a new strategy that supports input-input relationship constraints and demonstrates the correctness of the strategy as well as the effectiveness of test data with input-input relationship.

Implementation of artificial bee colony algorithm for T-way testing

Literature evidences have demonstrated the effectiveness of the sampled t-way test suite for defect detection in software testing. The main task in implementing t-way testing strategies is constructing best possibility test case. There are several methods that have been proposed but none of them can be claimed to be the best result because t-way are considered as NP-hard problem. In this paper, the concept of artificial bee colony (ABC) will be proposed as the idea in constructing t-way testing test suit. Due to the limitation of classical optimization in solving larger scale combinatorial and highly non-linear problems, researchers are moving to employ the intelligent behaviour of swarm known as Swarm Intelligence. ABC was introduced by Karaboga in 2005 is an optimization algorithm based on honey bee swarm which has been applied in solving real world application.

Test case generation for event driven systems using 4-way input test strategy

Since 1990s Combinatorial explosion is a crucial topic in software testing domain. In this domain the exhaustive input interaction test is not practical solution to test all software configurations for generating test cases. T-way test strategy provides a solution to overcome the problem. To date, most of the research on designing methods focus on combinations of input parameter values, where the T-way test strategy is applied to generate final test cases. However the methods are not applicable for event driven systems, where inputs are tested in the ordered form. In this paper a new method is proposed on T-way strategy to generate test cases using 4-way input interaction technique. The generated test cases are compared with the existing 4-way strategies. Result shows that in most cases the proposed method is able to generate efficient test cases compared with existing 4-way methods.

Effectiveness comparison of partially executed t-way test suite based generated by existing strategies

Consuming 40 to 50 percent of software development cost, software testing is one of the most resource consuming activities in software development lifecycle. To ensure an acceptable level of quality and reliability of a typical software product, it is desirable to test every possible combination of input data under various configurations. Due to combinatorial explosion problem, considering all exhaustive testing is practically impossible. Resource constraints, costing factors as well as strict time-to-market deadlines are amongst the main factors that inhibit such consideration. Earlier work suggests that sampling strategy (i.e. based on t-way parameter interaction or called as t-way testing) can be effective to reduce number of test cases without effecting the fault detection capability. However, for a very large system, even t-way strategy will produce a large test suite that need to be executed. In the end, only part of the planned test suite can be executed in order to meet the aforementioned constrai...