One of the most serious cyberattacks where investigators are concerned about a solution is phishing. Phishing is a technique used by attackers to entice end users and steal their private data. To obtain personal information, attackers deceive Internet users by impersonating a legitimate website. This can also be accomplished by posing as legitimate companies or businesses in emails. Phishing successfully exploits several vulnerabilities, and there is no one sizefits-all solution to protect users from all vulnerabilities. To minimize the harm caused by phishing it is necessary to be discovered as soon as possible. There are several methods for detecting phishing depending on a whitelist, black-list, content-based, URL-based, Visual Similarity, And machine-learning. In this study, was proposed hybrid ensemble approach based on the combination of Random Forest, AdaBoost and xgboost to detect phishing websites into two phases. Firstly, were individually performed each model. Secondly, would combine models and analyze hybrid ensemble model to get the best combination of ensemble classifiers that works robust on phishing website attacks. The proposed approach evaluated using an imbalanced dataset, with a higher percentage of legitimate URLs than phishing URLs. The dataset is used to train and test each classification model and hybrid ensemble model. The experimental results show that the proposed hybrid ensemble approach achieved an accuracy of 95.23%, According to the findings, the proposed hybrid ensemble can detect phishing websites with high accuracy. In the future, will suggest that ensemble classifiers can be combined with Deep Learning techniques to create hybrid models used for phishing website detection. Keywords: Phishing, Phishing websites, Legitimate, Hybrid ensemble