Indoor robots, essential components of modern cyber-physical systems, face growing threats from disruptive attacks that jeopardise their functionality and data security. Tan Le, Van Le, and Sachin Shetty from Old Dominion University, present a new framework for detecting malicious software in these robots, combining the power of quantum computing with deep learning techniques. This innovative approach achieves remarkably high detection accuracy, up to 95. 2%, even while protecting sensitive privacy information, and crucially, operates without relying on pre-defined settings or continuous data collection. The research demonstrates a significant step forward in building trustworthy artificial intelligence for autonomous systems, offering robust, interpretable, and stable performance in challenging real-world environments.
Machine learning secures cyber-physical systems
Research focuses on enhancing the security, performance, and capabilities of cyber-physical systems, including robots, autonomous systems, and the Internet of Things. A key goal is to address vulnerabilities to cyberattacks, particularly those targeting systems that determine location and navigation. Scientists are leveraging machine learning, and especially quantum machine learning, to achieve these improvements by optimizing resource allocation and improving overall performance. Quantum machine learning utilizes algorithms like variational quantum algorithms and quantum neural networks, well-suited for current quantum computers.
These techniques, combined with deep learning and reinforcement learning, allow for complex pattern recognition and data analysis, and are integrated with robotics frameworks like the Robot Operating System and machine learning libraries like Scikit-learn. This work addresses critical challenges in cybersecurity, resource management, smart healthcare, and cognitive radio networks. Important considerations include the limitations of current quantum computers, the challenges of training quantum machine learning models, and the need for large amounts of data. Researchers are focused on ensuring that these models are explainable, secure, and can be successfully integrated into existing systems, while operating on resource-constrained devices and protecting sensitive data. Ultimately, this research aims to build more secure, intelligent, and efficient cyber-physical systems.
Quantum Malware Detection with Privacy Preservation
Scientists have developed a privacy-aware malware detection framework for indoor robotic systems, protecting them from Denial of Service attacks that disrupt location tracking and control. This innovative system combines deep learning with quantum-enhanced computing to accurately detect malicious activity while preserving data privacy. The system operates without relying on pre-defined thresholds or persistent beacon data, enabling scalable deployment in challenging environments. The methodology employs quantum feature encoding, harnessing quantum superposition to achieve exponential speedup in feature space exploration, a significant advancement over traditional artificial intelligence models.
Experiments utilize UWB-based systems, commonly used in mobile robotics, but acknowledge their vulnerability to spoofing and jamming. Benchmarking demonstrates robust generalization and resilience against training instability through modular circuit design. Numerical results reveal that hybrid quantum models outperform classical baselines in malware detection tasks, achieving up to 95. 238% accuracy and F1 scores exceeding 0. 95 in adversarial settings. This achievement demonstrates the potential of quantum-enhanced artificial intelligence to secure cyber-physical systems, particularly where real-time decision-making and privacy preservation are paramount, and establishes a foundation for secure, autonomous operation in indoor robotics.
Privacy-Aware Malware Detection with Quantum Computing
This work presents a breakthrough in cybersecurity for indoor robotic systems, specifically addressing the threat of Denial of Service attacks that compromise critical functions like localization and control. Scientists developed a privacy-aware malware detection framework that integrates deep learning with hybrid quantum computing to achieve unprecedented accuracy and resilience. The system operates without relying on pre-defined thresholds or persistent beacon data, enabling scalable deployment even when facing adversarial conditions. Experiments reveal the framework achieves up to 95. 2% detection accuracy under privacy-constrained conditions, demonstrating a significant advancement over traditional methods.
Furthermore, the team measured F1 scores exceeding 0. 95 in adversarial settings, confirming the system’s robust performance when subjected to malicious interference. A key innovation lies in the use of a hybrid quantum-classical system, benefiting from exponential speedup in feature space exploration via quantum superposition and improved generalization in noisy environments. The research team successfully implemented a full quantum neural network pipeline, leveraging remote access to Noisy Intermediate-Scale Quantum devices for telemetry encoding and inference. This allows robotic systems to offload malware detection to quantum co-processors while maintaining real-time control locally. Crucially, the framework incorporates explainability overlays via QuXAI, providing transparent insights into the decision-making process, and establishes a foundation for secure, autonomous operation in indoor robotics.
Quantum Malware Detection for Robotic Systems
This research presents a novel malware detection framework designed to protect indoor robotic systems from Denial of Service attacks within cyber-physical systems. The team successfully integrates quantum-enhanced feature encoding with deep learning techniques, achieving up to 95. 2% detection accuracy while preserving data privacy. Unlike conventional intrusion detection systems, this approach operates without relying on pre-defined thresholds or persistent beacon data, which allows for scalable deployment in dynamic and privacy-sensitive environments. The resulting hybrid quantum-classical neural network demonstrates robust performance, interpretability, and resilience against signal manipulation, even when features are suppressed or incomplete. The architecture is designed to operate effectively on current quantum hardware, requiring only four qubits and exhibiting efficient gradient estimation, and is readily portable to cloud-based quantum infrastructure. Further research will continue to emphasize reproducibility, modularity, explainability, and ethical considerations, contributing to the development of trustworthy artificial intelligence for robotics and autonomous systems.
👉 More information
🗞 Privacy-Aware Framework of Robust Malware Detection in Indoor Robots: Hybrid Quantum Computing and Deep Neural Networks
🧠ArXiv: https://arxiv.org/abs/2510.13136