Download Full-Text PDF Cite this Publication
- Open Access
- Total Downloads : 69
- Authors : Ashritha Jain R, Mrs. Mangala Kini, Chaithra Kulal, Deekshitha S
- Paper ID : IJERTCONV7IS08034
- Volume & Issue : RTESIT – 2019 (VOLUME 7 – ISSUE 08)
- Published (First Online): 13-06-2019
- ISSN (Online) : 2278-0181
- Publisher Name : IJERT
- License: This work is licensed under a Creative Commons Attribution 4.0 International License
A Review Paper on Detection of Phishing Websites using Machine Learning
Ashritha Jain R
Computer Science and Engineering Alvas Institute of Engineering and Technology
Mrs. Mangala Kini
Assistant Professor Computer Science and Engineering
Alvas Institute of Engineering and Technology Moodbidri, India
Computer Science and Engineering Alvas Institute of Engineering and Technology
Computer Science and Engineering Alvas Institute of Engineering and Technology
Abstract Phishing is the fraudulent attempt to obtain sensitive information of individuals or organization such as usernames, passwords and credit card details by disguising as trustworthy entity in a electronic communication. Phishing attack causes serious threats to users privacy and security. The purpose of this study is to presents an overview about various phishing attacks and various techniques to protect the information. It also includes the discussion of Extreme Learning Machine (ELM) based classification for 30 features including phishing websites data in UC Irvine Machine Learning Repository database.
Keywords- Phishing, Extreme Learning Machine.
Internet has become an important part of our life to obtain, spread information in social media. While Mobile Social Networks enrich peoples lives, it also creates some security issues . In one of the previous studies the author defined phishing as a type of semantic attack in an online environment, where the victims are sent spoofed emails which essentially deceive them into providing confidential data such as account numbers, passwords and other personal information to the attacker . To understand what phishing does, we must know the different types of phishing.
Types of Phishing Attacks:
Numerous different types of phishing attacks have now been identified. Some of the more prevalent are listed below.
Deceptive phishing is the most common type of phishing. In this case, an attacker attempts to obtain confidential information from the victims. Attackers use the information to steal money or to launch other attacks.
Spear Phishing targets specific individuals instead of a wide group of people. Attackers often research their victims on social media and other sites. That way, they can customize their communications and appear more authentic.
When attackers go after a big fish like CEO, its called Whaling. These attackers often spend considerable time profiling the target to find the opportune moment and means of stealing login credentials.
Similar to phishing, Pharming sends users to a fraudulent website that appears to be legitimate. However, in this case, victims do not even have to click a malicious link to
be taken to the bogus site.
This paper is structured as follows: Section II discusses works and different methods presented in the literature for phishing detection. Section III introduces the proposed methodology that can be implemented to predict the phishing website accurately. Finally, the investigation gap that provides more scope to study about the phishing detection is in Section IV. Conclusion is given in Section V.
Simulator was developed to check the content and determine whether the related message contained phishing elements. Tianrui Peng et al.  proposed and named a system as SEAHound processes a document, one sentence at a time and returns tree if the document contains a social engineering attack. It focuses on the natural language text contained in the attack, performing semantic analysis of the text to detect malicious intent. This approach performs a semantic analysis of the text transmitted by the attacker to verify the appropriateness of each sentence. Jhen-Hao Li et al.  proposed an approach, called PhishBox to effectively collect phishing data and generates models for phishing validation and detection. It integrates phishing websites collection, detection and validation into an online tool which monitors the blacklisted phishing sites, validates and detects them in real-time. Naghmeh Moradpoor et al.  proposes a neural network-based model for detection and classification of phishing emails. It uses real benign emails from SpamAssassin dataset and real phishing emails from Phishcorpus dataset. Python and MATLAB is used to measure the accuracy, true-positive rate, false positive-rate, network performance, and error histogram. R.Aravindhan et al.  proposed a list based anti phishing approach, which has two types 1.Black list 2.White list. In black list some online databases such as phish tank provides list of phishing websites. In white list the user manually builds a white list by adding the trusted website to the white list. In heuristics based anti phishing approach the characteristics are determined such that it reflects the nature of the website accurately, machine learning techniques is used to find the phishing. Mustafa Aydin et al.  proposed a classification algorithm for phishing website detection by extracting websites' URL features and analyzing subset based feature selection methods. It implements feature extraction and selection methods for the detection of phishing websites. The extracted features about the URL of the pages and composed feature matrix are categorized into five different analyses as Alpha-numeric Character Analysis, Keyword Analysis, Security Analysis, Domain Identity Analysis and Rank Based Analysis. Most of these features are the textual properties of the URL itself and others based on third parties services. Samuel Marchal et al.  presents PhishStorm, an automated phishing detection system that can analyze in real time any URL in order to identify potential phishing sites. Phish storm is proposed as an automated real-time URL phishingness rating system to protect users against phishing content. PhishStorm provides phishingness score for URL and can act as a Website reputation rating system.
There are many algorithms that are used to detect the phishing websites accurately. Few of them are discussed in this section that can be used to classify the URL as legitimate or phished. The publicly available phishing websites data set from the UCI machine learning repository can be used for training and testing. The featurs of the dataset is used to predict the result.
Different algorithms that can be used to detect the phishing websites are:
Artificial Neural Networks (ANN)
An artificial neural network (ANN), inspired from biological neural networks, is a set of interconnected nodes (neurons). Each connection between nodes is typically assigned weights. The network learns by adjusting the weights, in the learning phase for correct prediction process. ANNs were considered less suitable for data mining due to their poor interpretability and long training times. However, their advantages include ability to classify patterns on which they have not been trained and high tolerance for noisy data.
K-Nearest Neighbour (k-NN)
Learning for k-NN classifiers occurs by analogy, that is, by comparing the test tuple to similar training tuples. These are distance-based comparisons that intrinsically assign equal weights to each attribute; therefore, accuracy could be poor when noisy or irrelevant data is presented. However, methods of editing and pruning have been introduced to solve the problem of useless and noisy data tuples respectively. The training tuples are described by n attributes. Each tuple represents a point in an n-dimensional space. The good value for the number of neighbors can be determined experimentally.
Support Vector Machine (SVM)
Support vector machines (SVMs) are used for the classification of both linear and nonlinear data. In short, when given an original training data, the algorithm uses a nonlinear mapping to transform it into a higher dimension. In this dimension, a linear optimal hyper plane is searched, to keep the data of any two classes separate. SVMs can be used for classification and numeric prediction as well. The simplest form of SVM is a two-class problem, where the classes are linearly separable. For a 2-D problem, a straight line can be drawn to separate the classes, in fact, multiple lines could be drawn.
Random Forests (RF)
Random Forests can be built in tandem with random attribute selection using bagging. Random Forests follow an ensemble approach to learning, that is a divide and conquer approach for improving performance. In a simple decision tree, the input or test in added at the top and it traverses down the tree, ending up in smaller subsets. In a random forest, the ensemble mechanism combines various random subsets of trees. The input/test traverses through all the trees. The result is calculated based on average or weighted average of the individual results, or the voting majority in case of categorical data. The accuracy of a random forest depends on a measure of the dependence between the classifier and the strength of the individual classiers and they improve the problem of over fitting of the decision trees.
The components for detection and classification of phishing websites include the discussion on thirty distinct attributes of websites. They are as follows:
Address Bar based Features
Using the IP address
If IP address is used instead of domain name in the URL e.g. 220.127.116.11 the user can almost be sure someone is trying to steal his personal information.
Long URL to hide the Suspicious Part
Phishers can use long URL to hide the doubtful part in the address bar.
Using URL shortening services TinyURL
URL shortening is a method on the World Wide Web in which a URL may be made considerably smaller in length and still lead to the required webpage.
URLs having @ symbol
Using @ symbol in the URL leads the browser to ignore everything preceding the @ symbol and the real address often follows the @ symbol.
Redirecting using //
The existence of // within the URL path means that the user will be redirected to another website.
Adding Prefix or Suffix Separated by (-) to the Domain The dash symbol is rarely used in legitimate URLs. Phishers tend to add prefixes or suffixes separated by (-) to the domain name so that users feel that they are dealing with a legitimate webpage.
Sub Domain and Multi Sub Domains
Let us assume we have the following link: http://www.hud.ac.uk/students/. A domain name might include the country-code top-level domains (ccTLD).
HTTPs (Hyper Text Transfer Protocol with Secure Sockets Layer)
The existence of HTTPS is very important in giving the impression of website legitimacy, but this is clearly not enough.
Domain Registration Length
Based on the fact that a phishing website lives for a short period of time, we believe that trustworthy domains are regularly paid for several years in advance. In our dataset, we find that the longest fraudulent domains have been used for one year only.
A favicon is a graphic image (icon) associated with a specific webpage.
Using Non-Standard Port
This feature is useful in validating if a particular service is up or down on a specific server.
The existence of HTTPS Token in the Domain Part of the URL
The phishers may add the HTTPS token to the domain part of a URL in order to trick users.
Abnormal Based Features
Request URL examines whether the external objects contained within a webpage such as images, videos and sounds are loaded from another domain.
URL of Anchor
An anchor is an element defined by the <a> tag. This feature is treated exactly as Request URL.
Links in <meta>, <Script> and <Link> tags
Given that our investigation covers all angles likely to be used in the webpage source code, we find that it is common for legitimate websites to use <Meta> tags to offer metadata about the HTML document; <Script> tags to create a client side script; and <Link> tags to retrieve other web resources.
It is expected that these tags are linked to the same domain of the webpage.
Server From Handler(SFH)
SFHs that contain an empty string or about:blank are considered doubtful because an action should be taken upon the submitted information.
Submitting Information to Email
Web form allows a user to submit his personal information that is directed to a server for processing. A phisher might redirect the users information to his personal email.
This feature can be extracted from WHOIS database. For a legitimate website, identity is typically part of its URL.
The fine line that distinguishes phishing websites from legitimate ones is how many times a website has been redirected. Status Bar Customization
Disabling Right Click
Using Pop-Up Window
It is unusual to find a legitimate website asking users to submit their personal information through a pop-up window.
IFrame is an HTML tag used to display an additional webpage into one that is currently shown.
Domain Based Features
Age of Domain
This feature can be extracted from WHOIS database. Most phishing websites live for a short period of time. By reviewing our dataset, we find that the minimum age of the legitimate domain is 6 months.
For phishing websites, either the claimed identity is not recognized by the WHOIS database or no records founded for the hostname. If the DNS record is empty or not found then the website is classified as Phishing, otherwise it is classified as Legitimate.
This feature measuresthe popularity of the website by determining the number of visitors and the number of pages they visit.
PageRank is a value ranging from 0 to 1. PageRank aims to measure how important a webpage is on the Internet.
This feature examines whether a website is in Googles index or not. When a site is indexed by Google, it is displayed on search results.
Number of Links Pointing to Page
The number of links pointing to the webpage indicates its legitimacy level, even if some links are of the same domain.
Statistical-Reports Based Feature
Several parties such as PhishTank formulate numerous statistical reports on phishing websites at every given period of time; some are monthly and others are quarterly.
INVESTIGATION ON RESEARCH GAPS
So far we have understood that phishing is a specialized social engineering attack whereby the attacker very intelligently uses spoofed emails or websites to trick the victims into sharing their confidential and sensitive information. There is a need to understand the psychology of online consumers that whether they are concerned about the security issues when they are having the authority to change the security features. There are many academic literatures about security against phishing. However, there are a number of issues that concern the gap between academic literature and practical evidence.
A major research gap exists between research and the industry in terms of true positives. While academic and literary research essentially focuses on machine-learning and heuristics, assuming very good true positives, these true positives are sometimes high false positives. Hence, these heuristics are only reasonable enough to identify phishing sites that have not been encountered before. However, the industry primarily relies on blacklists for classification of phishing websites. But, the blacklists fail to generalize to the future unseen cases and are also potentially slow in responding to zero-hour attacks.
CONCLUSION AND FUTURE WORK
This paper has presented three important elements of the study, a theory of phishing crime, a review of anti- phishing technique offered by different research and investigation of the research gaps. Phishing will never be eliminated, but it is important to understand this crime before proposing any solution. Here, we have discussed about
different features of phishing attacks and different techniques to detect phishing websites.
The future work will be to get the research into the development of phishing detection system particularly against phishing websites since it is considered the most common way of attack. For more accurate results, instead of NaÃ¯ve Bayesian approach, we can use Artificial Neural Network or Random Forest Classifiers. This detection tool will help to protect users from phishing attacks in the non-secured environment too.
Yasin SÃ¶nmez, TÃ¼rker Tuncer, HÃ¼seyin GÃ¶kal and Engin Avci, Phishing Web Sites Features Classification Based on Extreme Learning Machine, International Symposium on Digital Forensic and Security (ISDFS), May 2018.
Anjum N. Shaikh, Antesar M. Shabut and M. A. Hossain, A literature review on Phishing Crime, Prevention Review and Investigation of gaps, 2016 10th International Conference on Software, Knowledge, Information Management & Applications.
Neda Abdelhamid, Fadi Thabtah and Hussein Abdel-jaber, Phishing Detection: A Recent Intelligent Machine Learning Comparison based on Models Content and Features, IEEE Int. Conf. on Intelligence and Security Informatics (ISI), pages 7277, 2017.
Muhemmet Baykara, Zahit Ziya GÃ¼rel, Detection of phishing attacks, 2018.
Tianrui Peng, Detecting Phishing Attacks Using Natural Language Processing and Machine Learning , 2018.
Jhen-Hao Li, PhishBox: An Approach for Phishing Validation and Detection, 2017.
Naghmeh Moradpoor, Employing Machine Learning Techniques for Detection and Classification of Phishing Emails, July 2017.
R.Aravindhan, Dr.R.Shanmugalakshmi, Certain Investigation on Web Application Security: Phishing Detection and Phishing Target Discovery, January 2016.
Mustafa Aydin, Nazife Baykal, Feature Extraction and Classification Phishing Websites Based on URL, 2015.
Samuel Marchal, Radu State, Jerome Francois, and Thomas Engel, PhishStorm: Detecting Phishing with Streaming Analytics, December 2014.