Towards a Quantum-classical Augmented Network

TL;DR

This work tackles efficient security in hybrid quantum-classical networks by introducing a privacy-labeling approach that decides when to use quantum encryption. It extends HTTP to a quantum-classical data frame (Q-HTTP) and evaluates four ML models—Logistic Regression, CNN, LSTM, and BiLSTM—for classifying messages as private or non-private using a synthesized Enron-based dataset. The results show that LSTM achieves the highest accuracy (~95%), with high recall for private content, enabling roughly half of messages to be protected quantumly while the rest use classical encryption, improving resource efficiency. The work lays groundwork for practical, resource-conscious quantum-augmented networks and points to future enhancements like attention mechanisms and cross-modal privacy detection.

Abstract

In the past decade, several small-scale quantum key distribution networks have been established. However, the deployment of large-scale quantum networks depends on the development of quantum repeaters, quantum channels, quantum memories, and quantum network protocols. To improve the security of existing networks and adopt currently feasible quantum technologies, the next step is to augment classical networks with quantum devices, properties, and phenomena. To achieve this, we propose a change in the structure of the HTTP protocol such that it can carry both quantum and classical payload. This work lays the foundation for dividing one single network packet into classical and quantum payloads depending on the privacy needs. We implement logistic regression, CNN, LSTM, and BiLSTM models to classify the privacy label for outgoing communications. This enables reduced utilization of quantum resources allowing for a more efficient secure quantum network design. Experimental results using the proposed methods are presented.

Figures (11)

• Figure 1: The proposed modified HTTP data-frame structure for a quantum-classical augmented network.
• Figure 2: METHODOLOGY: Introduction of an ML model that parses a message and assigns a privacy label to each message, depending on the presence of any private information. Once any such quantum encryption passes through the quantum gateway and reaches the end user, this NLP model should be able to reassess the full message before the user can read it. Blue arrows indicates quantum payload, and red arrows indicates classical payload. We notice the presence of both arrows in Stage II, which is to signify that in a communication scenario, multiple packets would be transmitted. Red arrows shows the path of the packets with just classically encrypted information, and blue arrows shows the path taken by packets with quantum encrypted information.
• Figure 3: Single-Fiber Q-HTTP Architecture. Alice’s ML Privacy Classifier (L7) determines if quantum security (QKD) is required (label=1). Messages flow downward through Alice's layers (L7 to L1), pass through the optical network fiber (via WDM $\lambda_c + \lambda_q$), and flow upward through Bob's layers (L1 to L7). The "Sender" and "Receiver" groups encapsulate the respective layers.
• Figure 4: PCA Visualization of Email Embeddings.
• Figure 5: t-SNE Visualization of Email Embeddings.