Multi-Factor Phishing Attack Detection and Prevention with Support Vector Machine
Download 23.07 Kb.
|
- Navigate this page:
- ABSTRACT
- Keywords— Alexa Web Rank, Classification, Page trust value, Phishing attacks, Phishing determination. INTRODUCTION
- LITERATURE SURVEY Nguyen, Luong Anh Tuan et. al. [1] has proposed an efficient approach for phishing detection using single-layer neural network.
- Abdul Ghafoor Abbasi et. al. [2] have worked on w eb Contents Protection, Secure Execution and Authorized Distribution.
- Bai Zhongying et. al. [3] has proposed the w ebpage Encryption Based on Polymorphic Javascript Algorithm.
- Sarasan, Sandra et. al. [4] have worked on detection and Prevention of Web Application Security Attacks.
- Johari, Rahul et. al. [5] has conducted a investigation on web application vulnerabilities (sqlia, xss) exploitation and security engine for sql injection.
- FINDINGS OF THE LITERATURE REVIEW
 Multi-Factor Phishing Attack Detection and Prevention with Support Vector Machine
Pardeep Singh1, Amandeep Sharma2, Amrit Kaur3 1Research Scholar, Computer Engineering Department, UCoE, Punjabi University, Patiala, India Pardeep.sekhon13@gmail.com 2Research Scholar, Computer Engineering Department, UCoE, Punjabi University, Patiala, India Sharma00.aman@gmail.com 3Assistant Professor, Computer Engineering Department, UCoE, Punjabi University, Patiala, India Amrit.tiet@gmail.com ABSTRACT The phishing attacks are launched for the stealing the information from the website users by channelizing the users to their fake links created as the duplicates of the original websites. The user login information hacking is done by sending the emails pointing towards the fake links. The trusted web sources are the secure sources which does not impose any security threat while the users login over the portals. In this paper, the security model has been proposed to protect against the phishing attacks. The phishing attacks are protected by analyzing the source of origin of the target sources. The phishing attacks are generalized in the multi-hop link forward dataset, which is collected after the analysis of the origination of the web sources. The proposed reads the integrity factors of the sources from the global sources such as Alexa Web Rank, Google Page Rank and Google Trust Factor for the evaluation of the trust levels of the target web pages. If the page trust factor is computed lower than the significant value, the any of the security information is denied and the web page is prohibited. The proposed model is intended to improve the results in the terms of hardened security and time complexity. Keywords— Alexa Web Rank, Classification, Page trust value, Phishing attacks, Phishing determination.
Websites and Web applications are progressively more used to transmit security crucial services that increase the target for different security attacks like phishing attack, SQL Injection etc. interaction of Many web applications with back-end DBMS which stores the users sensitive information (like monetary, health), compromise with web security upshot in breaching the large amount of users information which leads to ethically to brutal economical loss and legal consequences. From [9] violation at Verizon illustrates that they now reign ultimate in both the quantity of breaches and the amount of data compromised. The web includes the number of different protocols to create communication between two ends with great security implanted individually by the protocols. Some of them are web server, HTTP protocol and server-side application development technologies (such as CGI, PHP, ASP, JAVA), client-side technologies (like JavaScript, Flash) via web browser (e.g. Firefox, Google chrome, internet explorer). Web application developed by using some scripting language and hosted over such a convoluted infrastructure confronts innate challenges and inconsistencies among them. Current all the widely-used web applications which are developed and tested over respective frameworks offers limited security resolution support. So web application security requires more efforts and tools from the attacks which could be impractical for computational market stress and for people devoid of adequate security skills or consciousness. As a consequence, a high percentage of net applications that are deployed on the web are enclosing to security vulnerabilities. According to a survey by the net Application Security Consortium, approximately forty ninth of the net applications being evaluated contain vulnerabilities of high threat level and over thirteen of the websites will be compromised utterly mechanically [10]. A recent report [11] expose that over 80% of the websites on the net have minimum solitary serious security vulnerability. The Internet has turn out to be an vital part of the modern world. It proposes a diversity of resources to business and individuals and is employed in nearly each form of administration and commerce. It has revolutionized conventional media and facilitated flawless, instantaneous communication worldwide. As more and more users accept mobile computing via the employ of smart phones and tablets, Internet priorities, products, and services have changed to congregate demand and obtain advantage of new technology, and they will persist to do so in the future. According to numerous studies, the majority of the financial and government organizations have extended their online services to their customers. For instance, in 2011, 83% of Americans and 85% of Europeans frequently shopped online (Fortune Magazine, 2011). As a effect, a new sort of electronic offense, called phishing, has shaped. Phishing is considered as online illegal trick of stealing users’ private information by sending email to users and influencing them to visit a fake-site that looks alike to original site. Phishing has been causing financial loss all over the world. In United States, phishing caused 3.2 billion costs in 2007. So, the anti-phishing problem is an imperative problem in modern society. Lots of studies have been proposed to identify phishing, like black list approach, heuristic-based approach, etc. However, there is even not a sufficient method. In this paper, we proposed a new method to identify the phishing sites that focuses on the features of URL (Primary Domain, Sub Domain, Path Domain) and the ranking of site (Page Rank, Alexa Rank, Alex Reputation). Then, a proposed single-layer neural network is a system which decreases the error and increases the performance.
Nguyen, Luong Anh Tuan et. al. [1] has proposed an efficient approach for phishing detection using single-layer neural network. As Blacklist based techniques cannot notice the phishing sites that are not in the blacklist database, in heuristic based approach, the weight of heuristic is depend on the training dataset. So, a novel phishing detection technique is proposed and maintained over a website and the weights of the heuristic are resultant according to the necessity to prevent from the outside attack. This will reduce the chances of attack and can make a robust system. Abdul Ghafoor Abbasi et. al. [2] have worked on web Contents Protection, Secure Execution and Authorized Distribution. This paper describes the structural design and implementation of an inclusive system which guards the Web contents from the diverse web attacks. Here the proposed system offers Web content protection: Source code files stored on web server would be in encrypted form, Secure Execution. Encrypted source code files would be decrypted, processed and re-encrypted on demand generated by another page, Authorized Distribution : A request generated would be checked for integrity of its source by some secret code or its url prior to processing it. Bai Zhongying et. al. [3] has proposed the webpage Encryption Based on Polymorphic Javascript Algorithm. Internet and intranet uses the HTML based web pages in order to create computer network. As WebPages load in browser the source code of HTML and JavaScript is also attained. By this any person can acquire the source code effortlessly. So, protection of web pages and its contents are very important. Authors proposed a security model for HTML by using the polymorphic JavaScript based encryption method. This encryption method assists to shield Java Scripts from polymorphic viruses. It also employs a phishing attack classification, permutation and check digit that amplifies security levels. Sarasan, Sandra et. al. [4] have worked on detection and Prevention of Web Application Security Attacks. Authors proposed a system to prevent from multiple security attacks on SQL injection over web application and Cross Site Scripting. Each one of the solutions has their peculiar strengths and weaknesses, and the developers have to elect the solutions according to their software development requirements. The impact of such attacks can be seen by the unavoidable occurrence of web applications for health care, banking, government and private administration, and so on. Here they confer some of the suspicious measures which can be deployed in web applications to avert the SQL Injection and XSS attacks. Johari, Rahul et. al. [5] has conducted a investigation on web application vulnerabilities (sqlia, xss) exploitation and security engine for sql injection. This paper comprises different types of attacks description like Cross site scripting attack, SQL injection vulnerabilities, and prevention techniques. Authors also include the possible prevention steps to be taken to prevent from the SQL Injection attacks. From this paper it is concluded that SQL Injection and Cross-site Scripting (XSS) Attacks are most influential and easiest attack methods on the Internet Application. This study provides a survey of recent techniques for shielding against SQL injection and XSS Exploits.
The existing model is heuristic-technique, which means it is capable of producing the accurate results on the given data consisted of one-type phishing attacks dataset. The proposed model can be improved to work with multi-type and multi-dimensional dataset. The existing model utilizes the Alexa Rank, Page Rank and Alexa Reputation for the trust factor analysis of the website. Whereas a new website may be seen on the lower level of Alexa popularity and page rank, which may falsely report the website as the phishing website. The existing model utilizes the four local features which contains the domain information, primary domain information, sub domain information and path domain information. The proposed model can utilize more local factors to analyze the trust and performance of the target websites.
For the phishing attack classification, we conduct a research of literature survey of existing models. In the literature study, we will examine the existing phishing attack classification models in details to obtain the research gaps. Literature study will guide towards getting the new solution for effectual development of robust phishing attack classification applications. It also becomes pretty significant to conduct a comprehensive literature review regarding the existing single layer neural network based phishing attack classification model and encoding based phishing attack classification applications and their performance parameters. The research work will be completed in both hypothetical and practical paradigm. A methodical post-completion performance examining survey would be performed to explore the efficiency of the existing model.
The existing model utilizes the neural network model for the purpose of automatic classification for the detection of phishing attacks. The existing model can be enhanced by using the support Vector machine classifier. The support Vector machine classification model can be added with more explanation and accuracy for the phishing model detection in the case of multidimensional and multivariate data. Also the support Vector machine may reduce the accuracy of the whole phishing attack mining system, which may be fixed using the strong and balanced pre-processing data. The proposed model can be also used with other popular and effective classifiers such as k-nearest neighbor or linear regression. The proposed model is expected to solve the problem of the accuracy, time complexity, etc. REFERENCES
Directory: uploaded files -> docs docs -> International Journal of Enhanced Research Publications, issn: 2319-7471 docs -> Review on Wireless Sensor Network Techniques for Security in Railways docs -> Design and Implementation of Home Energy Management System Based on gsm network docs -> Mobile and Wireless Networks Abstract a mobile phone docs -> A new nis distributed Model using jadex framework Mohammadreza Nami, Maryam Beikmohammadloo docs -> International Journal of Enhanced Research in Science, Technology & Engineering issn: 2319-7463, Vol. 5 Issue 2, February-2016 docs -> Design and Implementation of Intelligent Agent Shell H. M. Mohamed docs -> Android application docs -> Efficient Generalized Forensics Framework for extraction and documentation of evidence from mobile devices Download 23.07 Kb. Share with your friends:
|