Selective forwarding attacks pose significant risks to the security of Internet of Things (IoT) networks, necessitating effective detection measures. This paper presents a Multilayer Perceptron (MLP)-based approach to enhance IoT security by identifying and mitigating selective forwarding attacks with high accuracy. Using the Cooja Simulator, we emulate IoT environments to acquire extensive data on network attributes essential for MLP analysis. Key steps, including data normalization, handling of missing values, feature selection, and hyperparameter optimization, are applied to refine the model’s detection capability. Our empirical evaluation, measured by accuracy, precision, recall, and F1 score, demonstrates the model’s effectiveness in distinguishing between normal and attack scenarios. This work contributes to the field by proposing a scalable and adaptive MLP framework suitable for practical IoT applications, underscoring its relevance given the evolving landscape of IoT security threats.