HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability Detection
As fifth-generation (5G) and next-generation communication systems advance and find widespread application in critical infrastructures, the importance of vulnerability detection becomes increasingly critical. The growing complexity of these systems necessitates rigorous testing and analysis, with st...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Transactions on Quantum Engineering |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10716796/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832586888852013056 |
|---|---|
| author | Yifeng Peng Xinyi Li Zhiding Liang Ying Wang |
| author_facet | Yifeng Peng Xinyi Li Zhiding Liang Ying Wang |
| author_sort | Yifeng Peng |
| collection | DOAJ |
| description | As fifth-generation (5G) and next-generation communication systems advance and find widespread application in critical infrastructures, the importance of vulnerability detection becomes increasingly critical. The growing complexity of these systems necessitates rigorous testing and analysis, with stringent requirements for both accuracy and speed. In this article, we present a state-of-the-art supervised hybrid quantum neural network named HyQ2 for vulnerability detection in next-generation wireless communication systems. The proposed HyQ2 is integrated with graph-embedded and quantum variational circuits to validate and detect vulnerabilities from the 5G system's state transitions based on graphs extracted from log files. We address the limitations of classical machine learning models in processing the intrinsic linkage relationships of high-dimensional data. These models often suffer from dead neurons and excessively large outputs caused by the unbounded range of the rectified linear unit (ReLU) activation function. We propose the HyQ2 method to overcome these challenges, which constructs quantum neurons by selecting random neurons' outputs from a classical neural network. These quantum neurons are then utilized to capture more complex relationships, effectively limiting the ReLU output. Using only two qubits, our validation results demonstrate that HyQ2 outperforms traditional classical machine learning models in vulnerability detection. The small and compact variational circuit of HyQ2 minimizes the noise and errors in the measurement. Our results demonstrate that HyQ2 achieves a high area under the curve (AUC) value of 0.9708 and an accuracy of 95.91%. To test the model's performance in quantum noise environments, we simulate quantum noise by adding bit flipping, phase flipping, amplitude damping, and depolarizing noise. The results show that the prediction accuracy and receiver operating characteristic AUC value fluctuate around 0.2%, indicating HyQ2’s robustness in noisy quantum environments. In addition, the noise resilience and robustness of the HyQ2 algorithm were substantiated through experiments on the IBM quantum machine with only a 0.2% decrease compared to the simulation results. |
| format | Article |
| id | doaj-art-ce49dc3e14a84962be843723c77bb792 |
| institution | Kabale University |
| issn | 2689-1808 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Transactions on Quantum Engineering |
| spelling | doaj-art-ce49dc3e14a84962be843723c77bb7922025-01-25T00:03:49ZengIEEEIEEE Transactions on Quantum Engineering2689-18082024-01-01511910.1109/TQE.2024.348128010716796HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability DetectionYifeng Peng0https://orcid.org/0009-0007-3306-9417Xinyi Li1https://orcid.org/0009-0000-8413-1603Zhiding Liang2Ying Wang3https://orcid.org/0000-0002-9004-7253School of Systems and Enterprises, Stevens Institute of Technology, Hoboken, NJ, USAPratt School of Engineering, Duke University, Durham, NC, USADepartment of Computer Science, Rensselaer Polytechnic Institute, Troy, NY, USASchool of Systems and Enterprises, Stevens Institute of Technology, Hoboken, NJ, USAAs fifth-generation (5G) and next-generation communication systems advance and find widespread application in critical infrastructures, the importance of vulnerability detection becomes increasingly critical. The growing complexity of these systems necessitates rigorous testing and analysis, with stringent requirements for both accuracy and speed. In this article, we present a state-of-the-art supervised hybrid quantum neural network named HyQ2 for vulnerability detection in next-generation wireless communication systems. The proposed HyQ2 is integrated with graph-embedded and quantum variational circuits to validate and detect vulnerabilities from the 5G system's state transitions based on graphs extracted from log files. We address the limitations of classical machine learning models in processing the intrinsic linkage relationships of high-dimensional data. These models often suffer from dead neurons and excessively large outputs caused by the unbounded range of the rectified linear unit (ReLU) activation function. We propose the HyQ2 method to overcome these challenges, which constructs quantum neurons by selecting random neurons' outputs from a classical neural network. These quantum neurons are then utilized to capture more complex relationships, effectively limiting the ReLU output. Using only two qubits, our validation results demonstrate that HyQ2 outperforms traditional classical machine learning models in vulnerability detection. The small and compact variational circuit of HyQ2 minimizes the noise and errors in the measurement. Our results demonstrate that HyQ2 achieves a high area under the curve (AUC) value of 0.9708 and an accuracy of 95.91%. To test the model's performance in quantum noise environments, we simulate quantum noise by adding bit flipping, phase flipping, amplitude damping, and depolarizing noise. The results show that the prediction accuracy and receiver operating characteristic AUC value fluctuate around 0.2%, indicating HyQ2’s robustness in noisy quantum environments. In addition, the noise resilience and robustness of the HyQ2 algorithm were substantiated through experiments on the IBM quantum machine with only a 0.2% decrease compared to the simulation results.https://ieeexplore.ieee.org/document/10716796/NextG vulnerability detectionquantum computingquantum neural networks (QNNs) |
| spellingShingle | Yifeng Peng Xinyi Li Zhiding Liang Ying Wang HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability Detection IEEE Transactions on Quantum Engineering NextG vulnerability detection quantum computing quantum neural networks (QNNs) |
| title | HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability Detection |
| title_full | HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability Detection |
| title_fullStr | HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability Detection |
| title_full_unstemmed | HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability Detection |
| title_short | HyQ2: A Hybrid Quantum Neural Network for NextG Vulnerability Detection |
| title_sort | hyq2 x2009 a x2009 hybrid x2009 quantum x2009 neural x2009 network for x2009 nextg x2009 vulnerability x2009 detection |
| topic | NextG vulnerability detection quantum computing quantum neural networks (QNNs) |
| url | https://ieeexplore.ieee.org/document/10716796/ |
| work_keys_str_mv | AT yifengpeng hyq2x2009ax2009hybridx2009quantumx2009neuralx2009networkforx2009nextgx2009vulnerabilityx2009detection AT xinyili hyq2x2009ax2009hybridx2009quantumx2009neuralx2009networkforx2009nextgx2009vulnerabilityx2009detection AT zhidingliang hyq2x2009ax2009hybridx2009quantumx2009neuralx2009networkforx2009nextgx2009vulnerabilityx2009detection AT yingwang hyq2x2009ax2009hybridx2009quantumx2009neuralx2009networkforx2009nextgx2009vulnerabilityx2009detection |