A Hybridizing Learning and Optimizing Routing Attack Detection Using Quantum Aware Adaptive Forest Framework

Abstract

The recent boom of Internet of Things (IoT) networks has led to a huge growth in the application of low-power and lossy networks (LLNs) based on the Routing Protocol for Low-Power and Lossy Networks (RPL). Despite the fact that RPL allows efficient routing and scalable communication between constrained IoT devices, it is quite susceptible to routing-based attacks on its operations, including blackhole, decrease rank, hello flooding, and version number attacks. These attacks exploit routing control messages and cause network topology disturbance, leading to poor communication reliability and data integrity. This study introduces a combined intrusion detection system called the Quantum-Aware Adaptive Forest Framework, which can be used to increase the security of the RPL-based IoT networks. The suggested system combines the adaptive feature intelligence, dynamic correlation analysis, and hybrid ensemble learning to identify anomalous routing behavior in real time. Moreover, a Quantum-Inspired Particle Swarm Optimization scheme is utilized to optimize the model hyperparameters and enhance the detection level and reduce the load on the computer. The model integrates Random Forest and Gradient Boosting network by a dual ensemble fusion mechanism, and then a meta-learning aggregator is used to perform the final classification. The experimental study of a large-scale dataset demonstrates that the proposed QA-AFF has a weighted F1-score of 99.44% and an accuracy of 99.47% that is much better than conventional optimization and classification thresholds. The proposed architecture is light and can be scaled to be extended to deployment in edge gateways and border routers in IoT infrastructures. The system strengthens the reliability, security and resilience of large-scale IoT environments by facilitating prompt routing attack detection and leading to higher classification performance.

Introduction

IoT networks are rapidly expanding and rely heavily on the RPL routing protocol, which is designed for low-power and resource-constrained devices. However, RPL has major security weaknesses, making it vulnerable to routing attacks such as Blackhole, Hello Flood, and Version Number attacks. These attacks can disrupt communication, cause packet loss, and drain energy, while traditional cryptographic solutions are often too heavy for IoT devices. This creates a need for lightweight and efficient intrusion detection systems (IDS).

Existing IDS approaches include rule-based methods and machine learning models like Random Forest, which perform well in detecting anomalies. However, most are either computationally expensive, limited to detecting a single attack type, or lack adaptability to dynamic IoT environments. This motivates the need for a more scalable, adaptive, and multi-attack detection framework.

The proposed solution, the Quantum-Aware Adaptive Forest Framework (QA-AFF), uses a large IoT intrusion dataset and reduces features to improve efficiency. It introduces a quantum-inspired optimization algorithm (QPSO-based) to tune a Random Forest model by selecting optimal parameters such as the number of trees and tree depth. The optimized model improves both accuracy and efficiency while remaining suitable for large-scale IoT data.

The final system builds an Adaptive Random Forest classifier (94 trees, depth 12) trained on over 1.4 million records. It classifies multiple attack types such as DDoS, DoS, Mirai, and Reconnaissance using majority voting across trees.

Conclusion

The fast-changing nature of the IoT ecosystems has resulted in more advanced, dynamic and scalable security designs to counter the more extreme network threats. This study presented a new framework called Quantum-Aware Adaptive Forest Framework (QA-AFF) a hybrid system of learning and optimization that was specifically created to be used in RPL-based IoT. The suggested framework balances high accuracy in forecasting and computing scalability by dynamically modifying the architectural hyperparameters of a Random Forest ensemble using a quantum-inspired swarm optimizer. The resulting system shows strong capacity to deal with large network data sets, successfully surmounting the drawbacks of static feature selection and conventional Newtonian optimisation tools. The use of experimental data on a dataset of 1.8 million rows highlights the effectiveness of the QA-AFF methodology, with an almost perfect F1-score of 99.44%. The comparative analysis has shown that the proposed framework not only performed better than traditional models such as SVM that had catastrophic complexity bottlenecks, but also outperformed the state-of-the-art gradient boosting frameworks such as LightGBM. Most distinctly, the quantum-tuned architecture was significantly more stable to classify devastating volumetric attacks like DDoS and Mirai with 100 % recall, and was much higher in recalling minority attack classes than the established GS-PSO approaches in previous literature. To sum up, the QA-AFF offers a very robust and dynamic solution to make IoT networks resource-starved and to avoid the resource-prohibitive overhead of sophisticated cryptographic protocols. Although the framework has a small training time trade-off with non-optimized models, its high structural stability and efficiency at inference time make it an optimal candidate to be deployed in the real world. Further research will be conducted on integration of feature-reduction methods and online learning modules in future efforts to further reduce training latency so that the framework can remain dynamic to the constantly-evolving environment of dynamic, multi-vector adversarial threats in heterogeneous IoT infrastructures.

References

[1] C. Yilmaz, S. Yilmaz, and S. Sen, Early Adaptive Intrusion Detection System for Mobile IoT Networks, IEEE Access, Vol. 13, pp. 205714–205732, 2025. C. Yilmaz, S. Yilmaz, and S. Sen, Early Adaptive Intrusion Detection System for Mobile IoT Networks, IEEE Access, Vol. 13, pp. 205714–205732, 2025. [2] U. Shahid, M. Z. Hussain, M. Z. Hasan, A. Haider, J. Ali, and J. Altaf, Hybrid Intrusion Detection System for RPL IoT Networks Using Machine Learning and Deep Learning, IEEE Access, Vol. 12, pp. 113099–113112, 2024. [3] S. Al-Sarawi, M. Anbar, B. A. Alabsi, M. A. Aladaileh, and S. D. A. Rihan, Passive Rule-Based Approach to Detect Sinkhole Attack in RPL-Based Internet of Things Networks, IEEE Access, Vol. 11, pp. 94081–94093, 2023. [4] M. Osman, J. He, F. M. M. Mokbal, N. Zhu, and S. Qureshi, ML-LGBM: A Machine Learning Model Based on Light Gradient Boosting Machine for the Detection of Version Number Attacks in RPL-Based Networks, IEEE Access, Vol. 9, pp. 83654–83665, 2021. [5] A. J. K. Q. Junior, E. T. Tchao, M. Al-Khalidi, A. S. Agbemenu, B. Yeboah-Akowuah, T. S. M. A. Adjaidoo, and E. Keelson, A Systematic Review on the Practicality of Poisoning Defenses in Federated IoT Systems, IEEE Access, Vol. 13, pp. 211109–211137, 2025. [6] E. Pagliari, L. Davoli, and G. Ferrari, Harnessing Communication Heterogeneity: Architectural Design, Analytical Modeling, and Performance Evaluation of an IoT Multi-Interface Gateway, IEEE Internet of Things Journal, Vol. 11, No. 5, pp. 8030–8051, 2024. [7] S. Lee, H. Choi, T. Kim, H. Park, and J. K. Choi, A Novel Energy-Conscious Access Point System with Cross-Layer Design in Wi-Fi Networks for Reliable IoT Services, IEEE Access, Vol. 10, pp. 61228–61248, 2022. [8] N. Nabeel, M. H. Habaebi, and M. D. R. Islam, Security Analysis of LNMNT-LightWeight Crypto Hash Function for IoT, IEEE Access, Vol. 9, pp. 165754–165765, 2021.

Copyright

Copyright © 2026 Rajdeep Saha, Anushka Rai, Animesh Kumar Choudhary, Dr. R Arun Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.