As phishing attacks become more common, strong detection mechanisms are urgently needed to protect against malicious Uniform Resource Locators (URLs). In our research, we propose a novel hybrid Tuna-Sea Horse Optimization Algorithm (hTSHOA), which combines two swarm-based optimization algorithms to optimize the algorithm through feature-defining URLs; this algorithm results in useful information and thus improves the feature selection process. In addition, the combined model uses a hyperparameter-tuned Support Vector Machine (SVM) in hybrid Ensemble model to determine if the URL is phishing or legitimate. Similarly, we regulate the hyperparameters of the SVM model using math based evolutionary metaheuristic technique referred to as the Arithmetic Optimization Algorithm (AOA), which enhances the model’s overall performance. We conclude that our approach works successfully with 98.28% accuracy at the ISCX-URL2016 dataset. This demonstrates that our generation is able to thwarting URL phishing assaults.