Internet of Things (IoT) intrusion detection is crucial for ensuring the security of interconnected devices in our digital world. With diverse devices communicating in complex networks, IoT environments face vulnerabilities such as dos attacks and unauthorized access. This paper proposes a novel Security Assessment FramEwork using Attention based Cnn-bigru for Iot Devices (SAFE-ACID) technique that preprocesses data through one-hot encoding, extracts features using Principal Component Analysis, and utilizes an Attention-based CNN-BiGRU model for intrusion detection. The study compares the proposed method with existing techniques using datasets like DS2OS, UNSW-NB15, and ToN_IoT, demonstrating superior presentation in terms of accuracy, F1 score, precision, detection rate, and security rate. According to the comparative analysis, the proposed technique’s detection rate is higher than the existing DRF-DBRF, HDA-IDS, and IDS-SIoEL techniques by 16.09%, 4.27%, and 6.9% respectively.