K. Nalini Sujantha Bel; I.Shatheesh Sam
Abstract
A novel feature extraction algorithm using Otsu’s Threshold (OT-features) on scrambled images and the Instantaneous Clustering (IC-CBIR) approach is proposed for Content-Based Image Retrieval in cloud computing. Images are stored in the cloud in an encrypted or scrambled form to preserve the privacy ...
Read More
A novel feature extraction algorithm using Otsu’s Threshold (OT-features) on scrambled images and the Instantaneous Clustering (IC-CBIR) approach is proposed for Content-Based Image Retrieval in cloud computing. Images are stored in the cloud in an encrypted or scrambled form to preserve the privacy content of the images. The proposed method extracts the features from the scrambled images using the Otsu’s threshold. Initially, the Otsu’s threshold is estimated from the scrambled image and based on this threshold the image is divided into two classes in the first iteration. Again, the new threshold values are estimated from two classes. The difference between the new threshold and the previous threshold gives two features. This process is repeated for number of iteration to obtain the complete OT-features of the scrambled image. This paper also proposes an instantaneous clustering approach (IC-CBIR) where the image is moved into a cluster as soon as the image is uploaded by the image owner. Therefore while retrieving the images, the images near to a particular cluster are matched instead of matching with a complete set of image features in the dataset which reduces the search time. The performance of the proposed algorithm is being tested using four different types of the dataset such as Corel 10K, Misc, Oxford flower, and INRIA Holidays dataset. The experimental evaluation reveals that the proposed method outperforms better than the traditional CBIR algorithm on encrypted images in terms of precision, time of search and time of index construction.
Dharmaraj Rajaram Patil; Jayantrao Patil
Abstract
Nowadays, malicious URLs are the common threat to the businesses, social networks, net-banking etc. However, malicious URLs deal with various Web attacks like phishing, spamming and malware distribution. Existing approaches have focused on binary detection i.e. either the URL is malicious or benign. ...
Read More
Nowadays, malicious URLs are the common threat to the businesses, social networks, net-banking etc. However, malicious URLs deal with various Web attacks like phishing, spamming and malware distribution. Existing approaches have focused on binary detection i.e. either the URL is malicious or benign. Very few literature is found which focused on the detection of malicious URLs and their attack types. Hence, it becomes necessary to know the attack type and adopt an effective countermeasure. This paper proposed a methodology to detect malicious URLs and the type of attacks based on multi-class classification. In this work, we proposed 42 new features of spam, phishing and malware URLs like URL Features, URL Source Features, Domain Name Features and Short URLs Features. These features are not considered in the earlier studies for malicious URLs detection and attack types identification. Binary and multi-class dataset is constructed using 49935 malicious and benign URLs. It consists of 26041 benign and 23894 malicious URLs containing 11297 malware,8976 phishing and 3621 spam URLs. To evaluate the proposed approach, state of the art supervised batch and online machine learning classifiers are used. Experiments are performed on the binary andmulti-class dataset using the aforementioned machine learning classifiers. It is found that, confidence weighted learning classifier achieved the best 98.44% average detection accuracy with 1.56% error-rate in the multi-class setting and 99.86% detection accuracy with negligible error-rate of 0.14% in binary setting using our proposed URL features.
M. Imani; Gh. A. Montazer
Abstract
The aim of phishing is tracing the users' s private information without their permission by designing a new website which mimics the trusted website. The specialists of information technology do not agree on a unique definition for the discriminative features that characterizes the phishing websites. ...
Read More
The aim of phishing is tracing the users' s private information without their permission by designing a new website which mimics the trusted website. The specialists of information technology do not agree on a unique definition for the discriminative features that characterizes the phishing websites. Therefore, the number of reliable training samples in phishing detection problems is limited. Moreover, among the available training samples, there are abnormal samples that cause classification error. For instance, it is possible that there are phishing samples with similar features to legitimate ones and vice versa. A supervised feature extraction method, called weighted feature line embedding, is proposed in this paper to solve these problems. The proposed method virtually generates training samples by utilizing the feature line metric. Hence, it can solve the small sample size problem. Moreover, by assigning appropriate weights to each pair of feature points, it corrects the undesirable quality of abnormal samples. The features extracted by our method improve the performance of phishing website detection specially by using small training sets.