Enhancing Phishing Website Detection via Feature Selection in URL-Based Analysis
Marwa-Aqa Abd Al Hussein Qasim, Nahla Abbas Flayh · Informatica · 2023
Detecting a phishing website accurately is crucial for ensuring the safety of online users, underscoring the importance of maintaining a secure digital environment. This research delves into the effectiveness of enhancing the detection of phishing websites through the application of a new dataset generation method. The method involves the transformation of a pure dataset obtained from Mendeley, by the utilization of regular expressions to extract the important features so that a detection process can be performed correctly with high performance. Based on the proposed features, we selected the best machine-learning algorithm.We performed a rigorous evaluation using Three prominent machine learning algorithms: Decision Trees, Support Vector Machines (SVM), and Random Forests, achieving 0.96% for Decision Tree Accuracy, 0.97% for SVM Accuracy, and 0.98% for Random Forest Accuracy.One of the critical contributions of this research is the deliberate selection of features. We have leveraged regular expressions to create a feature set that captures salient aspects of URLs and optimizes the algorithms' detection capabilities.This research has examined how feature selection affects the performance of each algorithm, highlighting its strengths and uncovering its weaknesses.Povzetek: glavni prispevek te raziskave je namerna izbira lastnosti. Izkoristili smo regularne izraze, da smo ustvarili nabor funkcij, ki zajame pomembne vidike URL-jev in optimizira zmožnosti zaznavanja algoritmov