The contemporary era has witnessed substantial growth in the Internet of Things (IoT), accompanied by an escalation in data volume due to the proliferation of Internet-connected devices. Unfortunately, this surge has attracted the attention of cybercriminals, who are now targeting IoT networks for malicious activities. Consequently, security and privacy concerns have emerged as the primary obstacles impeding the widespread adoption of IoT. While complete prevention of attacks on any system is not feasible, real-time detection of such attacks is crucial for effectively safeguarding IoT systems. This paper introduces a novel Intrusion Detection System (IDS) that employs supervised machine learning and bio-inspired algorithms to identify security anomalies in IoT networks. The IDS scrutinizes all dataset features to detect intrusive data and trains itself to predict potential network intrusions. However, some features may be extraneous or unrelated to the detection process, resulting in computational complexities and increased detection time. To tackle this issue, we implemented a feature selection process using a modified firefly algorithm to eliminate irrelevant features from the dataset and choose the optimal subset of features (reducing from 74 to 39). This optimization enhances the performance of the IDS by decreasing detection time and improving prediction accuracy. The obtained results affirm that the proposed intrusion detection system is capable of identifying real-world intrusions and can serve as an effective security solution for IoT systems.