A link checker phishing is a system that helps employees to identify if they are on a fraudulent webpage. This is important because phishing scams are becoming increasingly convincing and sophisticated to fool unsuspecting victims.
A variety of machine learning based techniques have been proposed to detect phishing websites. These techniques use various heuristic features like web traffic, search engine, WHOIS record and Page Rank to improve the detection accuracy. However, these heuristic features might be present in the benign websites too and they are third-party dependent. Furthermore, these heuristic features cannot detect the websites hosted on hacked servers. Thus, a new and light-weight technique is needed for phishing website detection that is adaptable at client side.
Real-Time Fraud Detection Using an API Service: How It Works
This study proposes a phishing detection model that uses a combination of RNN and LSTM to predict malicious webpages. The LSTM feature vectors are trained on a dataset that contains both phishing and benign URLs. The resulting model has excellent performance on the phishing URL dataset, with an FPR of 2.7% and a TPR of 98%. It also achieves comparable results on the Airline Twitter dataset.
A strong classifier is key to successful phishing detection. This is why this study utilizes a boosting method, XGBoost, on integrated features including URL character sequence, various hyperlinks information, login form features, textual content-based features and other related heuristic features. This approach improves the classification performance by a significant margin. This demonstrates that this is an effective approach to detecting phishing websites and provides an efficient way to mitigate cybercrime attacks.