Artificial Intelligence Models for Balancing Energy Consumption and Security in 5G Networks †
Abstract
1. Introduction
- Utilizing recurrent neural networks (RNNs) [6] for predictive energy management, forecasting energy usage trends based on network traffic patterns to enable proactive resource optimization.
- Integrating explainable AI (XAI) techniques [1] not only to enhance the interpretability and transparency of the security module’s decisions but also, as our results suggest, to potentially improve detection performance through better feature understanding and model refinement.
- Developing an adaptive resource allocation mechanism, potentially leveraging insights from device-to-device (D2D) communication strategies [7], that optimizes resource distribution (e.g., power, bandwidth) based on the outputs of the RNN and CNN/XAI modules, guided by a multi-objective function.
2. Related Works
2.1. Energy Optimization in 5G Networks
2.2. AI-Based Intrusion Detection Systems
2.3. Integration of Energy and Security Solutions
3. Methodology
3.1. Problem Formulation
- E(x) represents the total energy consumption [11], calculated as
- represents the overall security risk, which can be expressed as
3.2. Predictive Energy Management
- h_t is the hidden state at time t,
- W_h, W_x are weight matrices,
- the input vector (e.g., traffic data), and
- σ is the activation function.
3.3. Anomaly Detection via CNNs
- denotes the convolutional weights,
- represents the input feature map,
- signifies the bias term
3.4. Resource Allocation Optimization
- denotes the traffic intensity at node ,
- signifies the residual energy at node ,
- indicates the total power utilized by node ,
- represents the risk score derived from anomaly detection,
- serves as the weight for balancing energy and risk priority.
3.5. Clustering and Resource Allocation Algorithm
3.6. Workflow Diagrams
- Traffic Data Collection: Historical and real-time traffic data is gathered from various network nodes.
- Preprocessing: Raw traffic data is normalized and denoised for RNN input
- RNN Model: The RNN model works by revealing temporal dependencies in the traffic patterns and predicting future energy demands. RNNs use hidden states to store information from past time steps to improve future predictions.
- Energy Prediction: The RNN produces energy forecast signals from each node for proactive energy management.
- Optimized Power Allocation: Power resources are allocated to nodes according to their predicted energy needs, which helps reduce energy wastage without compromising performance.
- Network Traffic Collection: Traffic data is collated from multiple sources across networks, including packet headers and payload metadata.
- Feature Extraction: Relevant features like volume of traffic, frequency and behavior patterns of packets are extracted and these serve as an input to the CNN model.
- CNN Model: The model learns the behavior patterns from the traffic [20] data and identifies the anomalies. Then one convolutional layer will extract higher-order features, which are discriminative to normal or abnormal behaviors.
- Anomaly Detection: CNN classifies traffic as normal or anomalous and does so with high confidence, such that it could therefore catch the malicious entities [21] (such as distributed denial-of-service (DDoS) attacks, or intrusion attempts, etc.).
- Threat Mitigation: Upon anomaly [22] detection, the system implements mitigation tactics like IP blacklisting or node quarantine.
4. Results and Analysis
4.1. Performance Metrics
- Energy Efficiency: The improvement in energy efficiency [23] is computed as:
- Anomaly Detection Accuracy: Accuracy (Acc %) is computed as:
4.2. Simulation Setup
4.3. Comparative Results
4.4. Visualization
- CPU Utilization: Adjust dynamically according to predicted traffic intensity.
- Energy Usage: Nodes with lower traffic were allocated minimal energy, conserving resources.
- Bandwidth: Prioritized for nodes with high-risk scores to maintain security and QoS.
5. Practical Examples
5.1. China Mobile—AI for Base Station Energy Optimization
- Implementation: China Mobile implemented AI algorithms across 100,000 base stations to automatically adjust power settings based on traffic patterns [25].
- Results: A20% rise in energy savings during non-peak hours with QoS preservation.
- Relevance: Highlights the value of predictive analytics in managing energy, aligning with the framework’s RNN-based energy optimization.
5.2. Vodafone—Renewable Energy Integration
- Implementation: Vodafone converted some of its off-grid base stations to solar and wind energy-driven systems and applied AI for cooling systems [26].
- Results: A 15% decrease in operational energy costs, reduced greenhouse gases.
- Relevance: Shows how AI can foster sustainable energy solutions through efficient energy management.
5.3. SK Telecom—Dynamic Network Slicing
- Implementation: SK Telecom combined AI and network slicing to support urban and rural deployments of dynamic slicing based on real-time analysis of traffic and security requirements [27]
- Results: An 18% gain in energy efficiency in urban settings and 25% less energy use in rural ones.
- Relevance: Resonates with dynamic resource allocation strategies in the framework for optimally balancing energy and security.
5.4. Ericsson—Energy-Efficient Hardware
- Implementation: Energy efficient hardware solutions were implemented by Ericsson such as adaptive power amplifiers and low-loss antennas, in conjunction with better performance tracking using AI-based monitoring [28].
- Results: Improved network reliability at peak loads and launched initiatives to reduce idle power usage by 50%.
- Relevance: Affirms the significance of hardware–software synergy in cost savings and resilient operations.
5.5. Hardware Palo Alto Networks—AI-Driven Threat Detection
- Implementation: The implementation of CNNs has been seen in the solutions provided by Palo Alto Networks as part of 5G security, helping to identify and address potential threats, including DDoS attacks and data breaches [29].
- Results: Improved anomaly detection accuracy by 95% and drastically shortened incident response times.
- Relevance: Shows application of CNNs in anomaly detection which serve as the cybersecurity component of the framework
5.6. Huawei—Intelligent Threat Mitigation
- Implementation: Huawei developed an AI-enabled IDS based on a federated learning (FL) approach that can train threat detection models across 5G nodes while avoiding sharing raw data [30].
- Results: A 30% increase in detection accuracy for zero-day attacks, and better data privacy during model training.
- Relevance: Shows the importance of decoupled, decentralized AI modeling for secure 5G operations as an enhancer of frameworks.
5.7. IBM—AI-Orchestrated Security Automation
- Implementation: IBM used XAI-driven orchestration systems to automate 5G network security, embedding AI models to identify advanced persistent threats (APTs) [31].
- Results: A 40% reduction in response time to APTs, with highly transparent decision-making.
- Relevance: Validates the need for transparent and scalable security mechanisms, aligning with the proposed framework’s use of XAI.
6. Discussion
6.1. Hardware Integration of Renewable Energy
6.2. AI-Driven Network Slicing
6.3. Cross-Layer Optimization
6.4. Quantum Computing and AI
6.5. Autonomous Self-Healing Networks
6.6. Socio-Economic Implications
7. Conclusions
- Up to 65% in energy savings, proving it works by optimizing the resource and reducing the operational cost.
- A 99.7% accuracy rate for anomaly detection, enabling swift detection and resolution of threats.
- Transparency improvement via XAI, which is crucial for trust management in AI-driven systems and correlating with ethical considerations for the deployment of AI.
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
Appendix A
- Energy Prediction: The RNN model predicts consumption of energy based on historical traffic information [19], allowing for energy resources to be managed proactively.
- Anomaly Detection: The CNN-based IDS detects potential security threats in real-time, calculating the anomaly score of every network node [18].
- Optimization: An optimization score is computed for each node in accordance with the trade-off between energy efficiency and security risks, thus determining the allocation for resources with a score higher than a prominent threshold.
- XAI Validation: Explainable AI offers transparency in decisions made by the algorithm to allocate; hence analyst input can be [19] integrated into the decisions and decision-making can be reshaped on the basis of human interpretation and feedback.
- State Updates: The network state reflects current energy levels and risk metrics, allowing for constant optimization.
Algorithm A1: Adaptive Resource Allocation for Energy and Security |
Input: Network nodes N = {n1, n2, …, nn} Traffic data T = {t1, t2, …, tn} Energy thresholds {E_min, E_max} Security risk weights W = {w1, w2, …, wn} Model parameters: RNN_weights, CNN_weights, XAI_parameters Output: Optimized resource allocation and security configuration Step 1: Initialize parameters Initialize residual energy E_residual for each node Initialize resource allocation matrix A = 0 Initialize security risk scores R = 0 Step 2: Predict energy consumption For each node ni in N do: - Input historical traffic data ti to RNN - Predict energy consumption using: = RNN_predict(T) Step 3: Detect anomalies For each node ni in N do: - Input real-time traffic data to CNN - Compute anomaly score S(ni) using: Step 4: Optimize resource allocation For each node ni in N do: - Compute optimization score O(ni) using: - Allocate resources based on O(ni): > Threshold: Assign resources and update A(ni) Step 5: Validate decisions using XAI For each decision in A: - Use Explainable AI (XAI) to generate interpretability report - Adjust thresholds or allocations if required based on interpretability feedback Step 6: Update network state For each node ni in N: - Update residual energy: - Log energy and risk metrics Step 7: Repeat until performance criteria are met While global energy savings < Target or risk levels > Tolerable: - Recompute predictions and anomaly scores - Optimize resource allocation iteratively End |
References
- Szpilko, D.; Fernando, X.; Nica, E.; Budna, K.; Rzepka, A.; Lăzăroiu, G. Energy in Smart Cities: Technological Trends and Prospects. Energies 2024, 17, 6439. [Google Scholar] [CrossRef]
- Mirindi, D.; Sinkhonde, D.; Mirindi, F. An Advance Review of Urban-AI and Ethical Considerations. In Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Advances in Urban-AI (UrbanAI ‘24), Atlanta, GA, USA, 29 October–1 November 2024; Association for Computing Machinery: New York, NY, USA, 2024; pp. 24–33. [Google Scholar] [CrossRef]
- Hisyam Ng, H.A.; Mahmoodi, T. Machine Learning-Driven Dynamic Traffic Steering in 6G: A Novel Path Selection Scheme. Big Data Cogn. Comput. 2024, 8, 172. [Google Scholar] [CrossRef]
- Lifelo, Z.; Ding, J.; Ning, H.; Qurat-Ul-Ain; Dhelim, S. Artificial Intelligence-Enabled Metaverse for Sustainable Smart Cities: Technologies, Applications, Challenges, and Future Directions. Electronics 2024, 13, 4874. [Google Scholar] [CrossRef]
- Bagwari, A.; Logeshwaran, J.; Raja, M.; Devisivasankari, P.; Bagwari, J.; Rathi, V.; Saad, A.M.E. Intelligent Computational Model for Energy Efficiency and AI Automation of Network Devices in 5G Communication Environment. Tsinghua Sci. Technol. 2024, 29, 1728–1751. [Google Scholar] [CrossRef]
- Huang, J.E.; Chang, L.W.; Chin, H.H. 5G Metaverse in Education. Mobile Netw. Appl. 2024, 29, 947–960. [Google Scholar] [CrossRef]
- Zhu, X.; Wei, Y.; Lu, Y.; Zhao, M.; Yang, K.; Wu, S.; Zhang, H.; Wong, K.K.L. Comparative analysis of active contour and convolutional neural network in rapid left-ventricle volume quantification using echocardiographic imaging. Comput. Methods Programs Biomed. 2021, 199, 105914. [Google Scholar] [CrossRef]
- El Ansari, R.; El Bouhadioui, M.; Aboutafail, M.O.; Mejjad, N.; Jamil, H.; Jamal, E.; Rissouni, Y.; Zouiten, M.; Boutracheh, H.; Moumen, A. A review of Machine learning models and parameters for groundwater issues. In Proceedings of the NISS ‘23: Proceedings of the 6th International Conference on Networking, Intelligent Systems & Security, Larache, Morocco, 24–26 May 2023. [Google Scholar] [CrossRef]
- Chhabra, S.; Aiden, M.K.; Sabharwal, S.M.; Al-Asadi, M. 5G and 6G Technologies for Smart City. In Enabling Technologies for Effective Planning and Management in Sustainable Smart Cities; Ahad, M.A., Casalino, G., Bhushan, B., Eds.; Springer: Cham, Switherland, 2023. [Google Scholar] [CrossRef]
- Imanbayev, A.; Jakupov, A.; Valikhan, Y.; Odarchenko, R. Designing an Effective Network-Based Intrusion Detecting System for 5G Networks. CEUR Workshop Proc. 2024, 3800, 12–18. [Google Scholar]
- Ez-zahouani, B.; EL Kharki, O.; Kanga Idé, S.; Zouiten, M. Determination of Segmentation Parameters for Object-Based Remote Sensing Image Analysis from Conventional to Recent Approaches: A Review. Int. J. Geoinform. 2023, 19, 23–42. [Google Scholar] [CrossRef]
- Prasad, K.S.; Lydia, E.L.; Rajesh, M.V.; Radhika, K.; Ramesh, J.V.N.; Neelima, N.; Pokuri, S.R. Augmenting cybersecurity through attention based stacked autoencoder with optimization algorithm for detection and mitigation of attacks on IoT assisted networks. Sci. Rep. 2024, 14, 30833. [Google Scholar] [CrossRef]
- Sholeyan, A.E.; Rahatabad, F.N.; Setarehdan, S.K. Designing an Automatic Sleep Staging System Using Deep Convolutional Neural Network Fed by Nonlinear Dynamic Transformation. J. Med. Biol. Eng. 2023, 43, 11–21. [Google Scholar] [CrossRef]
- Almohaimeed, M.; Albalwy, F. Enhancing IoT Network Security Using Feature Selection for Intrusion Detection Systems. Appl. Sci. 2024, 14, 11966. [Google Scholar] [CrossRef]
- Gupta, R.; Singh, A. Machine learning models for anomaly detection in 5G networks. IEEE Trans. Netw. Serv. Manag. 2022, 19, 1021–1034. [Google Scholar]
- Farzaneh, B.; Shahriar, N.; Al Muktadir, A.H.; Towhid, M.S.; Khosravani, M.S. DTL-5G: Deep transfer learning-based DDoS attack detection in 5G and beyond networks. Comput. Commun. 2024, 228, 107927. [Google Scholar] [CrossRef]
- Senevirathna, T.; La, V.H.; Marcha, S.; Siniarski, B.; Liyanage, M.; Wang, S. A Survey on XAI for 5G and Beyond Security: Technical Aspects, Challenges and Research Directions. IEEE Commun. Surv. Tutor. 2025, 27, 941–973. [Google Scholar] [CrossRef]
- Zhang, Y.; Lin, C. Energy efficiency in 5G: Challenges and AI-based solutions. J. Wirel. Commun. Netw. 2020, 18, 789–805. [Google Scholar]
- Baghirov, E. A comprehensive investigation into robust malware detection with explainable AI. Cyber Secur. Appl. 2025, 3, 100072. [Google Scholar] [CrossRef]
- Chen, W.; Huang, P. AI for real-time threat detection in ultra-dense 5G networks. In Proceedings of the IEEE Global Communications Conference (GLOBECOM), Kuala Lumpur, Malaysia, 4–8 December 2023; pp. 512–523. [Google Scholar]
- Xu, R.; Zhang, X. AI strategies for scalable energy management in 5G smart grids. Elsevier Appl. Energy 2023, 349, 1–15. [Google Scholar]
- Tsourdinis, T.; Chatzistefanidis, I.; Makris, N.; Korakis, T.; Nikaein, N.; Fdida, S. Service-aware real-time slicing for virtualized beyond 5G networks. Comput. Netw. 2024, 247, 110445. [Google Scholar] [CrossRef]
- Haseeb, K.; Alruwaili, F.F.; Khan, A.; Alam, T.; Wafa, A.; Khan, A.R. AI Assisted Energy Optimized Sustainable Model for Secured Routing in Mobile Wireless Sensor Network. Mobile Netw. Appl. 2024, 29, 867–875. [Google Scholar] [CrossRef]
- Segura-Garcia, J.; Sturley, S.; Arevalillo-Herraez, M.; Alcaraz-Calero, J.M.; Felici-Castell, S.; Navarro-Camba, E.A. 5G AI-IoT System for Bird Species Monitoring and Song Classification. Sensors 2024, 24, 3687. [Google Scholar] [CrossRef]
- Solomon Arockiadass, S.A.M.; Thangavel, N.; Prakash A, J.F.; Raja, V.; Dhaigude, R. Enhanced 6G MIMO Signal Detection With Dynamic Sparse Region-Based Convolutional Neural Network With White Shark Optimization. Internet Technol. Lett. 2025, 8, e616. [Google Scholar] [CrossRef]
- China Mobile Limited. Annual Report 2023—Green Telecom Initiatives. Available online: https://www.chinamobileltd.com/en/ir/reports/ar2023.pdf (accessed on 1 June 2025).
- Vodafone Group Plc. Sustainability Report 2022. Available online: https://www.annualreports.com/HostedData/AnnualReportArchive/v/LSE_VOD_2022.pdf (accessed on 1 June 2025).
- SK Telecom. Energy Efficiency/Climate Change Report 2021. Available online: https://www.sktelecom.com/img/pds/persist_biz/2021/2021_CDP_Climate_Change_SK_Telecom.pdf (accessed on 1 June 2025).
- Sustainability and Corporate Responsibility Report 2022. Available online: https://www.responsibilityreports.com/HostedData/ResponsibilityReportArchive/e/NASDAQ_ERIC_2022.pdf (accessed on 1 June 2025).
- Palo Alto Networks Cybersecurity Insights Report, 2023. Available online: https://www.paloaltonetworks.com/resources/research/2023-unit-42-attack-surface-threat-report (accessed on 1 June 2025).
- Huawei Technologies Co.; China Mobile Communications Group Co. Green Management White Paper 2023. Available online: https://www.huawei.com/minisite/otf2023/file/Green_Management_White_Paper.pdf (accessed on 1 June 2025).
- BM Telecommunications Security Report, 2022. Available online: https://www.ibm.com/investor/att/pdf/IBM_Annual_Report_2022.pdf (accessed on 1 June 2025).
Metric | CNN (Baseline) | CNN + XAI | Proposed Framework |
---|---|---|---|
Energy Savings (%) | 50 | 55 | 65 |
Detection Accuracy (%) | 98.5 | 99.0 | 99.7 |
Latency (ms) | 12 | 11 | 10 |
Scenario | Energy Savings (%) | Detection Accuracy (%) |
---|---|---|
Normal Traffic | 60 | 99.0 |
High Traffic (DDoS) | 65 | 99.7 |
Mixed Traffic | 62 | 99.5 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Lazrek, H.; El Ferindi, H.; El Mouhtadi, M.; Zouiten, M.; Moumen, A. Artificial Intelligence Models for Balancing Energy Consumption and Security in 5G Networks. Eng. Proc. 2025, 112, 23. https://doi.org/10.3390/engproc2025112023
Lazrek H, El Ferindi H, El Mouhtadi M, Zouiten M, Moumen A. Artificial Intelligence Models for Balancing Energy Consumption and Security in 5G Networks. Engineering Proceedings. 2025; 112(1):23. https://doi.org/10.3390/engproc2025112023
Chicago/Turabian StyleLazrek, Hammad, Hassan El Ferindi, Meryam El Mouhtadi, Mohammed Zouiten, and Aniss Moumen. 2025. "Artificial Intelligence Models for Balancing Energy Consumption and Security in 5G Networks" Engineering Proceedings 112, no. 1: 23. https://doi.org/10.3390/engproc2025112023
APA StyleLazrek, H., El Ferindi, H., El Mouhtadi, M., Zouiten, M., & Moumen, A. (2025). Artificial Intelligence Models for Balancing Energy Consumption and Security in 5G Networks. Engineering Proceedings, 112(1), 23. https://doi.org/10.3390/engproc2025112023