As the number of devices connected are more in the Internet of Things (IoT), these devices create different data sharing, which leads to security and privacy issues. Current IoT cybersecurity solutions often lack accuracy, learning difficulties, and inaccurate costs. A novel DDoS attack detection framework is proposed based on hybrid binary grey wolf optimization (BGWO) metaheuristic algorithm and SHAP to solve these problems. A hybrid model of SHAP-BGWO is proposed to optimize feature selection. Machine learning tree classifiers are used to evaluate the performance on BoT-IoT dataset. The results of the above data, including accuracy, demonstrate the effectiveness of our SHAP-BGWO based framework in predicting DDoS attacks on IoT systems, providing solutions to improve IoT security. SHAP is combined with BGWO to achieve 99+% accuracy using AdaBoost, Random Forest and decision tree respectively.