Malicious Url Detection

Malicious Uniform Resource Locator/malicious websites are a high threat to cyber security. Malicious URLs host uninvited content like malware, spam, drive by downloads phishing, etc. Users become victims of scams like financial loss, thieving of personal data, malware installation, and causes losses of millions of dollars every year. There is a need to detect those threats in a very efficient and timely manner. Several studies have examined different techniques to handle the problem; the foremost used approach remains blacklisting. The most obstacle to using blacklist is that the difficulties in maintaining an up-to-date list of URLs. So here we proposed the Machine learning approach to detect the malicious URLs. We also discussed various methods for malicious URL detection, feature representations, and finally discussed various algorithms for the classification and feature extraction.

In this paper we look at the task of detecting URLs corresponding to infected web pages using Machine Learning and Natural Language Processing specific features. We show that these features render better performance than the previously used hand-crafted lexical features and render similar results to the more expensive host-based features. We also introduce a new adjacent task, that of identifying URLs pointing to the download of portable executable files, and show that our models perform very well on this task too.

Today the most important concern in the field of cyber security is finding the serious problems that make loss in secure information. It is mainly due to malicious URLs. Malicious URLs are generated daily. This URLs are having a short life span. Various techniques are used by researchers for detecting such threats in a timely manner. Blacklist method is famous among them. Researchers uses this blacklist method for easily identifying the harmful URLs. They are very simple and easy method. Due to their simplicity they are used as a traditional method for detecting such URLs. But this method suffers from many problems. The lack of ability in detecting newly generated malicious URLs is one of the main drawbacks of Blacklist method. Heuristic approach is also used for identifying some common attacks. It is an advanced technique of Blacklist method. But this method cannot be used for all type of attacks. So this method is used very shortly. For a good experience, the researchers introduce machine learning techniques. Machine Learning techniques go through several phases and detect the malicious URLs in an accurate manner. This method also gives the details about the false positive rate. This review paper studies the different phases such as feature extraction phase and feature representation phase of machine learning techniques for detecting malicious URLs. Different machine learning algorithms used for such detection is also discuss in this paper. And also gives a better understanding about the advantage of using machine learning over other techniques for detecting malicious URLs and problems it suffers.

The main aim of this project is to provide an easy and smooth browsing experience over the internet. This is to mainly optimize the user experience of surfing through various webpages. Users mainly get frustrated due to redirecting of page to annoying web-pages. To add to this, suppose the wi-fi , Ethernet or relative service's network is weak, this redirection causes a tedious task for the user to search or obtain the required result. Our project focuses on these drawbacks and will try to improve the user's browsing experience by blocking unwanted notification popups and redirection of website automatically. It also allows user to access the website only after user's approval to visit that website. It also keeps track of recently visited website with the help of cookies. This project helps user to protect their device by stopping them from entering malicious websites and also get rid of multiple popups

The phenomenon of phishing has now been a common threat, since many individuals and webpages have been observed to be attacked by phishers. The common purpose of phishing activities is to obtain user's personal information for illegitimate usage. Considering the growing intensity of the issue, this study is aimed at developing a new hybrid rule-based solution by incorporating six different algorithm models that may efficiently detect and control the phishing issue. The study incorporates 37 features extracted from six different methods including the black listed method, lexical and host method, content method, identity method, identity similarity method, visual similarity method, and behavioral method. Furthermore, comparative analysis was undertaken between different machine learning and deep learning models which includes CART (decision trees), SVM (support vector machines), or KNN (K-nearest neighbors) and deep learning models such as MLP (multilayer perceptron) and CNN (convolutional neural networks). Findings of the study indicated that the method was effective in analysing the URL stress through different viewpoints, leading towards the validity of the model. However, the highest accuracy level was obtained for deep learning with the given values of 97.945 for the CNN model and 93.216 for the MLP model, respectively. The study therefore concludes that the new hybrid solution must be implemented at a practical level to reduce phishing activities, due to its high efficiency and accuracy.

The trust and reputation models were introduced to restrain the impacts caused by rational but selfish
peers in P2P streaming systems. However, these models face with two major challenges from dishonest
feedback and strategic altering behaviors. To answer these challenges, we present a global trust model
based on network community, evaluation correlations, and punishment mechanism. We also propose a
two-layered overlay to provide the function of peers’ behaviors collection and malicious detection.
Furthermore, we analysis several security threats in P2P streaming systems, and discuss how to defend
with them by our trust mechanism. The simulation results show that our trust framework can successfully
filter out dishonest feedbacks by using correlation coefficients. It can effectively defend against the
security threats with good load balance as well.

Internet use is growing every day, accessing a website via its URL (Uniform Resource Locator) address is a daily task, but not all websites are benign to be accessed without any fear from malicious aims not matter where those websites are being accessed from (Web Browsers, e-mails body, chat application, SMS, VoIP) neither the nature of the operating system or the device. Our thesis aim is being able to detect the kind of websites that try to steal any user’s (normal users, communities, societies, laboratories, etc.) personal information like name, date of birth, e-mail, credentials, login and passwords from e-banking services for example or any other web services. Unlike traditional techniques that consists of penetrating data sources of web services providers by decrypting algorithms, the man idea of this kind of criminal activities is letting the victims give those informations unconsciously, by creating fake emails or websites that looks very similar of original ones and tell vi...

The phenomenon of phishing has now been a common threat, since many individuals and webpages have been observed to be attacked by phishers. The common purpose of phishing activities is to obtain user’s personal information for illegitimate usage. Considering the growing intensity of the issue, this study is aimed at developing a new hybrid rule-based solution by incorporating six different algorithm models that may efficiently detect and control the phishing issue. The study incorporates 37 features extracted from six different methods including the black listed method, lexical and host method, content method, identity method, identity similarity method, visual similarity method, and behavioral method. Furthermore, comparative analysis was undertaken between different machine learning and deep learning models which includes CART (decision trees), SVM (support vector machines), or KNN ( K -nearest neighbors) and deep learning models such as MLP (multilayer perceptron) and CNN (convol...