NetMamba is an efficient and linear-time state space model designed for network traffic classification. It addresses the challenges of model inefficiency and inadequate traffic representation in existing methods by using a unidirectional Mamba architecture instead of the Transformer. The model employs a comprehensive traffic representation scheme that retains valuable packet content while removing biases. Evaluation on six public datasets shows that NetMamba achieves nearly 99% accuracy, improves inference speed by up to 60 times, and demonstrates superior few-shot learning capabilities. The model's effectiveness, efficiency, and robustness are validated through extensive experiments, making it a promising solution for network traffic classification tasks.NetMamba is an efficient and linear-time state space model designed for network traffic classification. It addresses the challenges of model inefficiency and inadequate traffic representation in existing methods by using a unidirectional Mamba architecture instead of the Transformer. The model employs a comprehensive traffic representation scheme that retains valuable packet content while removing biases. Evaluation on six public datasets shows that NetMamba achieves nearly 99% accuracy, improves inference speed by up to 60 times, and demonstrates superior few-shot learning capabilities. The model's effectiveness, efficiency, and robustness are validated through extensive experiments, making it a promising solution for network traffic classification tasks.