Phishing Website Detection with Ensemble Learning Approach Using Artificial Neural Network and AdaBoost

Phishing is one of the most serious security threats. Most phishing attacks occur on online transaction websites such as banking, commercial businesses, e-commerce, and more. This type of attack is increasing every day. More than 90% of data breaches are done by phishing. In addition, ransomware viruses can also be delivered through phishing. Phishing websites have a similar appearance and URL to official and popular websites, making them very difficult to identify. Almost 75% of phishing websites use Secure Sockets Layer certificates (SSL) so the SSL protocol does not guarantee the website is legitimate. This research performs phishing detection with a URL-based approach, where URL data that has been extracted features will be analyzed and learned with machine learning techniques. Artificial Neural Network and AdaBoost are implemented to learn patterns in the data. To maximize the modeling results, some tests are also conducted on the ANN parameters. Based on the results of the ANN and AdaBoost implementation in detecting phishing websites, the hybrid ANN-AdaBoost model is the best with an accuracy of 98.60%, precision of 98.95%, and recall of 99.31%. This shows that the use of a combination technique of ANN and AdaBoost is the right choice to detect and increase accuracy and effectiveness in avoiding phishing threats.