Our everyday lives are getting increasingly connected with the Internet of Things, which presents serious worries about potential cybersecurity threats and the need for trustworthy solutions. Deep learning-based intrusion detection can increase the efficacy of the system. However, it is especially difficult to use these models in the Internet of Things. In this work, a lightweight intrusion detection model, using Feedforward Neural Network, has been developed. Model compression techniques, such as weight and neuron pruning, are applied to feedforward neural network. Performance evaluations of the baseline intrusion detection model and the proposed model have been conducted. Based on experimental results, the proposed model outperforms the baseline in terms of training time by an average of 93% while exhibiting very little degradation in performance.