The exponential growth of Internet of Things (IoT) devices has greatly expanded the potential targets for cybersecurity threats, with malware emerging as a serious issue. This paper presents an advanced framework for detecting and analyzing malware attacks on IoT networks, with a focus on optimizing resource allocation. Manipulating the RT-IoT2022 a private dataset, includes both benign and malicious network activities, offering a broad illustration of real-life situations is obtained from Kaggle, the study employs advanced techniques in data preprocessing, including cleaning, normalization, and feature extraction, followed by the partitioning of the dataset into training and testing subsets. The adaptation of the Light Gradient Boosting Machine (LightGBM) framework is central to the model's focus on detecting malware and analyzing its impact on IoT resource allocation, particularly in predicting shifts in resource usage patterns associated with various malware attacks. The enhanced IoT device features 4 Gigabytes (GB) of Random Access Memory (RAM) and a 2.0 GHz quad-core ARM Cortex- A72 processor, facilitating extensive testing and proving the model’s efficacy through advanced metrics. Results reveal promising outcomes, with accuracy rates surpassing 99.01%, coupled with high recall (98.98%), precision (99.01%), and F1-Score (99.56%) metrics, signifying the models’ robustness in identifying malicious activities. According to the comparison, the LightGBM model outperforms the Autoencoder in pattern detection in accuracy, recall, precision, and F1-score. It also shows that the LightGBM model is more efficient and effective than the quantized autoencoder-u8 (QAE) model in terms of resources, Central Processing Unit (CPU), memory, and time consumption.