Phishing Website Detection Model Construction Based on Improved IHHO-RF-SHAP Algorithm

With the rapid advancement of artificial intelligence technology, an increasing number of studies integrate artificial intelligence with optimization algorithms to address prediction, classification, and optimization problems. However, phishing websites bring great risks to network security and seriously threaten the property safety of users, and the construction of a reasonable and effective phishing website detection model can effectively improve network security. Therefore, this study proposes an improved Harris Hawk Optimization (IHHO, Improvement Harris Hawks Optimization) and couples the Random Forest model for phishing website detection, and finally gives an explain ability analysis. Firstly, an improved Harris Hawks Optimization according to chaotic mapping, random shrinkage index, and adaptive inertia weights is proposed, and the effectiveness of the algorithm is tested using five Benchmark functions; secondly, the IHHO algorithm is combined with the Random Forest to construct a parallel integration algorithm and apply it to the construction of phishing website detection model; finally, the global explanation of the model is provided by aggregating the SHAP values of multiple data points, (identifying the key drivers and the corresponding degree of influence) to improve the explainability of the algorithm and improve user's experience.