Feature extraction based ensemble stacking for combating cyber threat in phishing URLs

Cyber threat is a process of exploiting sensitive information by losing the golden triad (Confidentiality, Integrity, and Availability). The threats may include Trojans, Phishing, and DDOS attacks. Phishing works as a social engineering technique where in the malicious attacker attempts in such a way that the appearance to the user would seem legitimate thereby causing the user to fall in a trap which may lead to the compromising of the user's sensitive information by means of illegitimate URLs or links etc. According to previous records approximately 18% of businesses lost their financial information due to phishing attack. Phishing can exploit data such as credentials, personal data, medical, etc. In this work we proposed a feature extraction-based stacking model where the URL (uniform resource locators) is identified as a legitimate or illegitimate URL. In the process of feature extraction, the features are considered based on hostname _length, path_length, fd _length, tld_length, count of -, count of @, count of? count of %, count of., count of =, count of http, count of https, count of www, count of digits, count of letters, dir., Ip (v6 or v4). We trained the model with different machine learning algorithms for chosen the best models for stacking the models. We considered the best four machine learning algorithms to be stacked and a meta-layer for the final prediction. We compare the metrics such as accuracy, recall, precision, and F1-score with other state of art models.