Imblalanced Learning for Lightweight Message Queue Telemetry Transport Protocol Test Classification

In IoT research, the lightweight Message Queue Telemetry Transport (MQTT) protocol is widely used due to its efficient communication performance, but at the same time it is vulnerable to increasingly sophisticated cyberattack threats. Due to the data imbalance problem in the IoT MQTTset dataset, this poses a serious challenge to the accuracy of attack detection. To this end, this paper proposes a solution that combines conditional generative adversarial networks with a hybrid deep learning model, CNN-LSTM, to efficiently identify and classify attacks against the MQTT protocol. First, Protocol-constrained BCGAN is used to balance the MQTTset dataset, and the data imbalance problem is solved by generating high-quality minority class samples. Then, a hybrid deep learning model based on CNNLSTM with a joint dual-attention mechanism is designed to accurately distinguish normal traffic from abnormal traffic. The experimental results show that the proposed method significantly outperforms traditional methods in key metrics such as accuracy, recall and $\mathbf{F 1}$ score, especially in detecting minority class attacks.