This research is our final project. In our everyday life we manage personal risk and infer which situations may be dangerous and avoid them accordingly. However, it translates poorly to the context of the malicious URLs, there are few effective heuristics to differentiate safe URLs from dangerous ones. Internet criminals depend on the absence of such indicators to prey on their marks. A simple URL can cause a lot of damage. The potential harm is so great that malicious links are considered one of the biggest threats to the digital world. Some of the most popular cyber attacks use URLs for the attacks, such as C&C and phishing. Even though there are a lot of studies conducted about the detection of malicious URLs, there’s still a lot to discover especially about the weak spots of the defense mechanisms. Since machine learning has become one of the most prominent methods of malware detection, A robust feature selection mechanism is proposed that results in malicious domain detection models that are resistant to evasion attacks. This mechanism exhibits high performance based on empirical data. This paper is meant to help identify whether a link is malicious or benign. We rely on the thesis written by Nitay Hason “Robust Malicious URL Detection”.We have tried to find additional features to improve the accuracy of the URL detection, and improved the scores of the classifier. Furthermore, it introduces novel features that are robust to the adversary’s manipulation. Based on extensive evaluation of the different feature sets and commonly used classification models this paper show that models which are based on robust features are resistant to malicious perturbations, and at the same time useful for classifying non-manipulated data. Index Terms—Malware detection, Robust features, Domain