The rapid expansion of Internet of Things (IoT) devices across smart homes, healthcare, agriculture, and industrial systems has increased the importance of firmware security. Firmware acts as the operational core of IoT devices, and vulnerabilities within it can lead to unauthorized access, data leakage, device malfunction, and large-scale cyberattacks. This study presents an IoT Firmware Vulnerability Detection System that identifies security risks in firmware files using a hybrid approach that combines machine learning with analytical feature-based assessment. The proposed system accepts firmware in BIN, HEX, and ELF formats and processes them through a structured workflow that includes file normalization, architecture detection, binary feature extraction, opcode analysis, and classification. Key features such as entropy, opcode diversity, string count, byte variance, and binary size are extracted to represent the behavioral characteristics of the firmware. These features are then analyzed using a Random Forest model to estimate vulnerability probability. To improve reliability and interpretability, the machine learning output is combined with heuristic risk scoring to classify firmware as safe, suspicious, or vulnerable. The results show that the hybrid framework provides effective and understandable vulnerability assessment while supporting multiple firmware types and architectures. The system also includes a user-oriented interface that presents risk scores, feature-based visualizations, and backend analysis logs for better transparency. This research is significant because it offers a practical, scalable, and explainable solution for strengthening IoT firmware security, supporting researchers, developers, and cybersecurity professionals in the early detection of firmware-level threats.