Quantum-Inspired Matrix Product States for Ultra-Fast Parallel Network Traffic Classification in High-Speed Networks

The exponential growth of high-speed cryptographic protocols and ۶G network infrastructures has made traditional Deep Packet Inspection and legacy deep learning architectures largely inadequate for real-time traffic classification, primarily due to their excessive computational overhead and massive parameter footprints. To address this critical research gap, this study introduces a novel computational framework leveraging Matrix Product States (MPS) for ultra-fast, parallel network traffic anomaly detection and protocol identification. By translating normalized streaming packet header features into a high-dimensional Hilbert space via non-linear trigonometric feature mapping, our methodology completely bypasses the heavy matrix multiplications characteristic of conventional neural networks. Instead, it employs efficient tensor contraction mechanisms along a sequential chain, preserving complex temporal and structural flow dependencies while radically reducing structural complexity. Rigorous experimental evaluations conducted on extensive benchmark network traffic streams demonstrate that the proposed architecture achieves a flawless ۱۰۰% classification accuracy and an unprecedented inference throughput exceeding ۱.۲۳ million packets per second. Crucially, this elite performance is realized with a mere ۱,۸۴۰ trainable parameters, establishing an extraordinary paradigm shift in parameter-efficiency. In conclusion, the empirical findings confirm that quantum-inspired tensor networks offer a robust, lightweight, and highly scalable foundation for next-generation edge computing and high-throughput backbone network monitoring, successfully resolving the perennial trade-off between model accuracy and operational velocity

کلیدواژه ها:

Quantum-Inspired Computing ، Matrix Product States ، Network Traffic Classification ، Parallel Processing ، Tensor Networks