Computer Science And Technology - Research Publications

Permanent URI for this collectionhttps://kr.cup.edu.in/handle/32116/82

Browse

End date: 2020

Search Results

Now showing 1 - 10 of 66
  • No Thumbnail Available
    Item
    Defect prediction model of static code features for cross-company and cross-project software
    (Springer Science and Business Media B.V., 2018-12-06T00:00:00) Singh, Satwinder; Singla, Rozy
    Software project metrics are seen needless in software industries but they are useful when some unacceptable situations come in the project (Satapathy et al., Proceedings of the 48th annual convention of CSI, vol 2, 2013). Mainly the focus of various defect prediction studies is to build prediction models using the regional data available within the company. So companies maintain a data repository where data of their past projects can be stored which can be used for defect prediction in the future. However, many companies do not follow this practice. In software engineering, the crucial task is Defect prediction. In this paper, a binary defect prediction model was built and examined if there is any conclusion or not. This paper presents the assets of cross-company and within-company data against software defect prediction. Neural network approach has been used to prepare the model for defect prediction. Further, this paper compares the results of with-in and cross-company defect prediction models. To analyse the results for with-in company two versions of Firefox (i.e. 2.0 and 3.0) were considered; for cross project one version of Mozila Sea Monkey (1.0.1); for cross-company validation one version of LICQ were considered. Main focus of the study is to analyse the behavior or role of software metrics for acceptable level of defect prediction. � 2018, Bharati Vidyapeeth's Institute of Computer Applications and Management.
  • No Thumbnail Available
    Item
    A new binarization method for degraded document images
    (Springer Science and Business Media B.V., 2019-09-17T00:00:00) Rani, Usha; Kaur, Amandeep; Josan, Gurpreet
    The binarization of image is an important stage in any document analysis system such as OCR. It converts the colored or grayscale images into monochromatic form to reduce the computational complexity in the next stages. In old document images in the presence of degradations (ink bleed, stains, smear, non-uniform illumination, low contrast, etc.) the separation of foreground and background becomes a challenging task. Most of the existing binarization techniques can handle only a subset of these degradations. We present a simple binarization method for old document images. The experimental results confirm that the proposed technique gives good binarization results in the presence of various degradations. It computes the Laplacian of an image to separate the foreground. The subtracted Laplacian image is binarized using a global threshold. Finally, the postprocessing using morphological functions is applied. The results are compared in terms of F-measure, PSNR, time complexity, and OCR based evaluations which shows that our method outperforms existing techniques like Niblack, Sauvola, Gatos, Zhou, NICK, Singh, and Bataineh. � 2019, Bharati Vidyapeeth's Institute of Computer Applications and Management.
  • No Thumbnail Available
    Item
    Time series data analysis of stock price movement using machine learning techniques
    (Springer, 2020) Parray, I.R; Khurana, S.S; Kumar, M; Altalbe, A.A.
    Stock market also called as equity market is the aggregation of the sellers and buyers. It is concerned with the domain where the shares of various public listed companies are traded. For predicting the growth of economy, stock market acts as an index. Due to the nonlinear nature, the prediction of the stock market becomes a difficult task. But the application of various machine learning techniques has been becoming a powerful source for the prediction. These techniques employ historical data of the stocks for the training of machine learning algorithms and help in predicting their future behavior. The three machine learning algorithms used in this paper are support vector machine, perceptron, and logistic regression, for predicting the next day trend of the stocks. For the experiment, dataset from about fifty stocks of Indian National Stock Exchange�s NIFTY 50 index was taken, by collecting stock data from January 1, 2013, to December 31, 2018, and lastly by the calculation of some technical indicators. It is reported that the average accuracy for the prediction of the trend of fifty stocks obtained by support vector machine is 87.35%, perceptron is 75.88%, and logistic regression is 86.98%. Since the stock data are time series data, another dataset is prepared by reorganizing previous dataset into the supervised learning format which improves the accuracy of the prediction process which reported the results with support vector machine of 89.93%, perceptron of 76.68%, and logistic regression of 89.93%, respectively. � 2020, Springer-Verlag GmbH Germany, part of Springer Nature.
  • No Thumbnail Available
    Item
    Comparative Analysis of Salt and Pepper Removal Techniques for Binary Images
    (Springer, 2020) Rani, U; Kaur, A; Josan, G.
    Binarization is the most important step in the OCR system that converts the gray level or colored images into bi-level form. In the case of degraded images, results after binarization mostly contain noises. Salt and pepper noise of different sizes is the most prevalent noise in binary images. For the better results of OCR process, it is necessary to denoise image before proceeding to the next stage. This paper conducts experiments with different existing salt and pepper noise removal methods such as median filter-based techniques and kFill algorithm-based techniques for binary document images. The statistical measures, namely, PSNR, SSIM, and EPI are used to evaluate the performance. � 2020, Springer Nature Singapore Pte Ltd.
  • Thumbnail Image
    Item
    Modified Sauvola binarization for degraded document images
    (Elsevier, 2020) Kaur, A; Rani, U; Josan, G.S.
    The binarization of historical documents is a difficult job due to the presence of many degradations. Many existing local binarization techniques use certain manually adjusted parameters. The output of these techniques is much dependent on the value of these parameters. One of such parameters is window size which is kept fixed for the whole text image. The fixed window size will not be able to perform well for images having variable stroke widths and text sizes. The proposed binarization technique (Modified Sauvola) is the modification of state of art Sauvola's binarization technique. It automatically computes window size dynamically across the image pixel to pixel using the stroke width transform (SWT). This led to reduction in number of manually adjusted parameters. The results are compared with the nine existing techniques using the quantitative measures: FM, PSNR, NRM, MPM, and DRD. The results show that the proposed method outperforms existing methods for images having variable stroke widths and text sizes. - 2020 Elsevier Ltd
  • No Thumbnail Available
    Item
    Static, dynamic and intrinsic features based android malware detection using machine learning
    (Springer, 2020) Mantoo B.A.; Khurana S.S.
    Android is one of the smartest and advanced operating systems in the mobile phone market in the current era. The number of smartphone users based on the Android platform is rising swiftly which increases its popularity all over the world. The rising fame of this technology attracts everyone toward it and invites more number of hackers in Android platform. These hackers spread malicious application in the market and lead to the high chance of data leakage, financial loss and other damages. Therefore, malware detection techniques should be implemented to detect the malware smartly. Different techniques have been proposed using permission-based or system call-based approaches. In this paper, a hybrid approach of static, dynamic and intrinsic features based malware detection using k-nearest neighbors (k-NN) and logistic regression machine learning algorithms. The intrinsic feature contribution has also been evaluated. Furthermore, linear discriminant analysis technique has been implemented to evaluate the impact on the detection rate. The calculation uses a publicly available dataset of Androtrack. Based on the estimation results, both the k-nearest neighbors (k-NN) and logistic regression classifiers produced accuracy of 97.5%.
  • No Thumbnail Available
    Item
    Analysis of Wormhole Attack on AODV and DSR Protocols Over Live Network Data
    (Springer, 2020) Mishra H.K.; Mittal M.
    Wireless ad hoc networks due to their open deployment architecture, are highly exposed to many security compromising attacks. These attacks can cause a lot of damage to privacy, security, and robustness of networks. The wormhole attack is believed to be one of the malicious attacks to detect as it can be performed without breaching any key or breaking any cipher in any wireless ad hoc network. A wormhole attack form a tunnel in the network using two or more malicious nodes to replay the data stealthily from one malicious node to other malicious end nodes in same or different network. In this way, the ad hoc networks are exploited by the attacker by either using the flaws in protocol design or in network architecture. So, there is requirement of security methods to make MANET routing protocols thwarting wormhole attack. In this research work, the wormhole attack has been performed over AODV and DSR protocols using the real-time live data introduced in simulator. The prevention technique was noted to successfully handling the attack by restoring the performance of network and alleviates the effect of attack from the network.
  • No Thumbnail Available
    Item
    Clustering of tweets: A novel approach to label the unlabelled tweets
    (Springer, 2020) Jan T.G.
    Twitter is one of the fastest growing microblogging and online social networking site that enables users to send and receive messages in the form of tweets. Twitter is the trend of today for news analysis and discussions. That is why Twitter has become the main target of attackers and cybercriminals. These attackers not only hamper the security of Twitter but also destroy the whole trust people have on it. Hence, making Twitter platform impure by misusing it. Misuse can be in the form of hurtful gossips, cyberbullying, cyber harassment, spams, pornographic content, identity theft, common Web attacks like phishing and malware downloading, etc. Twitter world is growing fast and hence prone to spams. So, there is a need for spam detection on Twitter. Spam detection using supervised algorithms is wholly and solely based on the labelled dataset of Twitter. To label the datasets manually is costly, time-consuming and a challenging task. Also, these old labelled datasets are nowadays not available because of Twitter data publishing policies. So, there is a need to design an approach to label the tweets as spam and non-spam in order to overcome the effect of spam drift. In this paper, we downloaded the recent dataset of Twitter and prepared an unlabelled dataset of tweets from it. Later on, we applied the cluster-then-label approach to label the tweets as spam and non-spam. This labelled dataset can then be used for spam detection in Twitter and categorization of different types of spams.
  • No Thumbnail Available
    Item
    Investigating the role of code smells in preventive maintenance
    (University of Tehran, 2019) Reshi J.A.; Singh S.
    The quest for improving the software quality has given rise to various studies which focus on the enhancement of the quality of software through various processes. Code smells, which are indicators of the software quality have not been put to an extensive study for as to determine their role in the prediction of defects in the software. This study aims to investigate the role of code smells in prediction of non-faulty classes. We examine the Eclipse software with four versions (3.2, 3.3, 3.6, and 3.7) for metrics and smells. Further, different code smells, derived subjectively through iPlasma, are taken into conjugation and three efficient, but subjective models are developed to detect code smells on each of Random Forest, J48 and SVM machine learning algorithms. This model is then used to detect the absence of defects in the four Eclipse versions. The effect of balanced and unbalanced datasets is also examined for these four versions. The results suggest that the code smells can be a valuable feature in discriminating absence of defects in a software.
  • No Thumbnail Available
    Item
    Detection and prevention of de-authentication attack in real-time scenario
    (Blue Eyes Intelligence Engineering and Sciences Publication, 2019) Sharma S.; Mittal M.
    Wireless Local Area Network (WLAN) is an infrastructure network in which nodes are connected to a centralized system to provide Internet access to mobile users by radio waves. But WLANs are vulnerable to Medium Access Control (MAC) layer Denial of Service (DoS) attacks due to the susceptibility of the management frames. An attacker can spoof the MAC address of the legitimate client and perform de-authentication attack to disconnect WLANs users from the access point. Many free tools are available in Kali Linux Operating System (OS) by which this attack can be performed and cause a security threat to WLAN users. The consequences of de-authentication DoS attack are frequent disconnection from Internet, traffic redirection, man-in-the-middle attack, and congestion. Despite enormous efforts in combating de-authentication DoS attack in the past decade, this attack is still a serious threat to the security of the cyber world. Medium Access Control Spoof Detection and Prevention (MAC SDP) DoS algorithm performs detection and prevention of de-authentication attack caused by spoofing MAC address. This algorithm is modified to make it more immune to the de-authentication attack and implemented in real-time scenario. The results show that the proposed technique increases the packet flow rate by 20.36%, reduces the packet loss by 95.71%, and reduces the down time and recovery time by 0.39 sec and 0.9 sec respectively as compared to MAC SDP DoS algorithm.