Harsh Kumar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Harsh Kumar has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
Michael Devetsikiotis
University of New Mexico,Department of Electrical and Computer Engineering,USA
6 shared publications- Henok Berhanu Tsegaye
Department of Electrical and Computer Engineering, University of New Mexico, Albuquerque, NM 87131, USA
5 shared publications - Michael Devetsikiotis
Department of Electrical and Computer Engineering, University of New Mexico, Albuquerque, NM 87131, USA
5 shared publications Petro M. Tshakwanda
University of New Mexico,Department of Electrical and Computer Engineering,USA
5 shared publications- Petro Mushidi Tshakwanda
Department of Electrical and Computer Engineering, University of New Mexico, Albuquerque, NM 87131, USA
5 shared publications Sanjeev Kumar
Department of Electrical and Computer Engineering University of Texas Rio Grande Valley Edinburg TX USA
5 shared publications
Latest Publications
- Open Access
- 2 November 2026
- Published by MDPI AG in Electronics
- Vol. 15 (21) , 5086
Unified Blockchain Provenance and Bayesian Reputation for Additive Manufacturing Supply Chains
In additive manufacturing (AM) supply chains, systems that record provenance and score reputation as separate transactions let lifecycle records and assessments diverge. We present a Hyperledger Fabric chaincode that commits a provenance event, the Bayesian reputation update it implies and their link in one read–write set. An on-chain policy admits ratings only from counterparty organizations; reputation gates authorize future activities but never block recording outcomes. Under client crashes in the measured partial-write window, the unified design produced no divergence in 3900 trials (2900 with version 1.0, 1000 with version 1.1), while a client-orchestrated two-chaincode deployment diverged in all 1000 trials of two control batches. A same-channel InvokeChaincode arrangement also produced none in 1000, at 4.5 ms more endorsement and 11.7% lower throughput at 100 transactions in flight. On four desktops, open-loop load saturates near 12 tx/s, and a 12-h soak at 6 tx/s committed 98.7% of submissions with no partial record found. Ledger experiments cover gate refusal, failed outcomes, recovery and disputes; with client retry, 540 contended ratings each committed once. A simulation quantifies gate errors and shows that evaluation-time decay reacts slowly when active actors’ performance drops. The scores record submitted outcomes; they are not validated measures of performance.
- 2 November 2026
- Published by MDPI AG in Electronics
- Vol. 15 (21) , 5086
Open AccessUnified Blockchain Provenance and Bayesian Reputation for Additive Manufacturing Supply Chains
In additive manufacturing (AM) supply chains, systems that record provenance and score reputation as separate transactions let lifecycle records and assessments diverge. We present a Hyperledger Fabric chaincode that commits a provenance event, the Bayesian reputation update it implies and their link in one read–write set. An on-chain policy admits ratings only from counterparty organizations; reputation gates authorize future activities but never block recording outcomes. Under client crashes in the measured partial-write window, the unified design produced no divergence in 3900 trials (2900 with version 1.0, 1000 with version 1.1), while a client-orchestrated two-chaincode deployment diverged in all 1000 trials of two control batches. A same-channel InvokeChaincode arrangement also produced none in 1000, at 4.5 ms more endorsement and 11.7% lower throughput at 100 transactions in flight. On four desktops, open-loop load saturates near 12 tx/s, and a 12-h soak at 6 tx/s committed 98.7% of submissions with no partial record found. Ledger experiments cover gate refusal, failed outcomes, recovery and disputes; with client retry, 540 contended ratings each committed once. A simulation quantifies gate errors and shows that evaluation-time decay reacts slowly when active actors’ performance drops. The scores record submitted outcomes; they are not validated measures of performance.
- Open Access
- 10 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (20) , 4631
A Security-Aware System Architecture for Latency-Bounded Ontology-Driven Integration in Additive Manufacturing
Semantic integration in Additive Manufacturing (AM) supervision demands unifying heterogeneous sensor streams and production context under strict latency and data-integrity constraints, yet conventional semantic stacks suffer from unpredictable query latencies, unguarded SPARQL (SPARQL Protocol and Resource Description Framework [RDF] Query Language) endpoints, and prohibitive reasoning overheads. This paper presents Semantic DREAM (Distributed Resilient and Emergent intelligence-based Additive Manufacturing), a security-aware architecture co-designed over a Web Ontology Language 2 Rule Language (OWL 2 RL) modular substrate with three mechanisms: (1) an abstract syntax tree (AST)-guarded read-only SPARQL engine that blocks destructive updates and caches template results; (2) a multi-protocol mashup layer aligning live Message Queuing Telemetry Transport (MQTT), Representational State Transfer (REST), Structured Query Language (SQL), and file feeds through Sensor, Observation, Sample, and Actuator (SOSA) observation minting and Quantities, Units, Dimensions, and Types (QUDT) unit normalization; and (3) an ingestion-time Shapes Constraint Language (SHACL) validation pipeline that gates observations before graph commit, decoupling structural checks from reasoning. Deductive materialization is performed offline at startup and verified with ROBOT as an intentional design choice to eliminate reasoning overhead from the streaming path, so online monitoring queries evaluate directly over the asserted graph. Single-host, lab-scale evaluations show that cached queries achieve sub-15 ms response times for recurring monitoring, while the worst-case event-driven decision cycle (incurring a post-invalidation cache miss) completes in 110 ms, well within the 200 ms supervisory deadline with 100% defect interception on injected anomalies. All measurements are single-host and lab-scale; factory-network transit and recorded telemetry are not evaluated.
- 10 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (20) , 4631
Open AccessA Security-Aware System Architecture for Latency-Bounded Ontology-Driven Integration in Additive Manufacturing
Semantic integration in Additive Manufacturing (AM) supervision demands unifying heterogeneous sensor streams and production context under strict latency and data-integrity constraints, yet conventional semantic stacks suffer from unpredictable query latencies, unguarded SPARQL (SPARQL Protocol and Resource Description Framework [RDF] Query Language) endpoints, and prohibitive reasoning overheads. This paper presents Semantic DREAM (Distributed Resilient and Emergent intelligence-based Additive Manufacturing), a security-aware architecture co-designed over a Web Ontology Language 2 Rule Language (OWL 2 RL) modular substrate with three mechanisms: (1) an abstract syntax tree (AST)-guarded read-only SPARQL engine that blocks destructive updates and caches template results; (2) a multi-protocol mashup layer aligning live Message Queuing Telemetry Transport (MQTT), Representational State Transfer (REST), Structured Query Language (SQL), and file feeds through Sensor, Observation, Sample, and Actuator (SOSA) observation minting and Quantities, Units, Dimensions, and Types (QUDT) unit normalization; and (3) an ingestion-time Shapes Constraint Language (SHACL) validation pipeline that gates observations before graph commit, decoupling structural checks from reasoning. Deductive materialization is performed offline at startup and verified with ROBOT as an intentional design choice to eliminate reasoning overhead from the streaming path, so online monitoring queries evaluate directly over the asserted graph. Single-host, lab-scale evaluations show that cached queries achieve sub-15 ms response times for recurring monitoring, while the worst-case event-driven decision cycle (incurring a post-invalidation cache miss) completes in 110 ms, well within the 200 ms supervisory deadline with 100% defect interception on injected anomalies. All measurements are single-host and lab-scale; factory-network transit and recorded telemetry are not evaluated.
- Open Access
- 23 September 2026
- Published by IntechOpen in AI, Computer Science and Robotics Technology
- Vol. 5
Coverage-Aware Gradient-Boosted Intrusion Detection for Smart-Meter Distributed Denial-of-Service Resilience
Advanced metering infrastructure (AMI) relies on continuous communication among smart meters, gateways, routers, and utility monitoring systems, increasing exposure to distributed denial-of-service (DDoS) traffic. This study presents a system-level intrusion-detection methodology integrating coverage-aware sensor placement with leakage-controlled gradient-boosted classification. The contribution lies in combining monitoring visibility, training-only feature sanitization, duplicate and cross-partition overlap removal, false-positive-constrained threshold selection, repeated external testing, and a validated feature-budget mode. Seventeen CIC-DDoS2019 parquet files were analyzed. After removing conflicting labels, exact duplicate feature vectors, and 4,666 testing records overlapping the training partition, the final dataset contained 121,088 training records and 297,790 external testing records with 60 predictors. Across 10 paired random-seed runs, XGBoost achieved mean accuracy 0.97707, precision 0.99991, recall 0.97258, F1-score 0.98606, receiver operating characteristic area under the curve 0.99944, Matthews correlation coefficient 0.92450, and false-positive rate 0.00044. Random Forest achieved mean F1-score 0.98492. The paired F1 improvement was 0.00113 (95% confidence interval 0.00061–0.00166; one-sided Wilcoxon p = 0.00098), with McNemar testing also favoring XGBoost. Family-level analysis showed weaker recall for synchronize (SYN), User Datagram Protocol (UDP)-lag, and WebDDoS traffic. Independent RT-IoT2022 replication yielded mean XGBoost F1-score 0.99583, while a port/service-stripped sensitivity analysis yielded 0.93984. A CICIoT2023 stress test detected SYN, UDP, and unseen HTTP-flood traffic with recall of at least 0.9998. Results support a reproducible first-stage AMI communication-layer detector while confirming dataset-dependent attack-family detectability.
- 23 September 2026
- Published by IntechOpen in AI, Computer Science and Robotics Technology
- Vol. 5
Open AccessCoverage-Aware Gradient-Boosted Intrusion Detection for Smart-Meter Distributed Denial-of-Service Resilience
Advanced metering infrastructure (AMI) relies on continuous communication among smart meters, gateways, routers, and utility monitoring systems, increasing exposure to distributed denial-of-service (DDoS) traffic. This study presents a system-level intrusion-detection methodology integrating coverage-aware sensor placement with leakage-controlled gradient-boosted classification. The contribution lies in combining monitoring visibility, training-only feature sanitization, duplicate and cross-partition overlap removal, false-positive-constrained threshold selection, repeated external testing, and a validated feature-budget mode. Seventeen CIC-DDoS2019 parquet files were analyzed. After removing conflicting labels, exact duplicate feature vectors, and 4,666 testing records overlapping the training partition, the final dataset contained 121,088 training records and 297,790 external testing records with 60 predictors. Across 10 paired random-seed runs, XGBoost achieved mean accuracy 0.97707, precision 0.99991, recall 0.97258, F1-score 0.98606, receiver operating characteristic area under the curve 0.99944, Matthews correlation coefficient 0.92450, and false-positive rate 0.00044. Random Forest achieved mean F1-score 0.98492. The paired F1 improvement was 0.00113 (95% confidence interval 0.00061–0.00166; one-sided Wilcoxon p = 0.00098), with McNemar testing also favoring XGBoost. Family-level analysis showed weaker recall for synchronize (SYN), User Datagram Protocol (UDP)-lag, and WebDDoS traffic. Independent RT-IoT2022 replication yielded mean XGBoost F1-score 0.99583, while a port/service-stripped sensitivity analysis yielded 0.93984. A CICIoT2023 stress test detected SYN, UDP, and unseen HTTP-flood traffic with recall of at least 0.9998. Results support a reproducible first-stage AMI communication-layer detector while confirming dataset-dependent attack-family detectability.
- Open Access
- 27 May 2026
- Published by MDPI AG in Journal of Manufacturing and Materials Processing
- Vol. 10 (6) , 183
Addressing Data Scarcity in Additive Manufacturing Monitoring via Synthetic Data Generation and Meta Pseudo-Labeling for Foundational Layer-Wise Segmentation
Additive manufacturing (AM) monitoring is fundamentally constrained by the severe scarcity of annotated data for layer-wise segmentation. This paper addresses this bottleneck by introducing a scalable, high-fidelity synthetic data generation pipeline built on the Slice-100K dataset, capable of producing large volumes of layer-wise semantic segmentation masks. Through analysis of this large-scale synthetic data, we identify a systemic foreground–background class imbalance (1:24 ratio) inherent to AM monitoring, which causes standard Dice loss formulations to diverge catastrophically into a phenomenon we formalize as the “Dice Crash.” To effectively leverage large amounts of unlabeled data, we adapt the Meta Pseudo-Labeling (MPL) framework for industrial segmentation. We evaluate MPL’s true marginal utility by integrating it with both a standard U-Net and a robust state-of-the-art nnU-Net architecture. Experimental outputs show that while MPL yields substantial performance gains (+15.2%) on weak baselines, integrating it with an optimally configured strong baseline consistently improves segmentation accuracy and suppresses false foreground detections, thereby mitigating confirmation bias. These findings demonstrate that semi-supervised learning via continuous bilevel optimization offers a practical and robust enhancement to data-scarce additive manufacturing monitoring. Because any hidden defects in the topmost layer will be permanently buried by subsequent extrusion, this foundational layer-wise segmentation step is the most critical primitive of the monitoring pipeline.
- 27 May 2026
- Published by MDPI AG in Journal of Manufacturing and Materials Processing
- Vol. 10 (6) , 183
Open AccessAddressing Data Scarcity in Additive Manufacturing Monitoring via Synthetic Data Generation and Meta Pseudo-Labeling for Foundational Layer-Wise Segmentation
Additive manufacturing (AM) monitoring is fundamentally constrained by the severe scarcity of annotated data for layer-wise segmentation. This paper addresses this bottleneck by introducing a scalable, high-fidelity synthetic data generation pipeline built on the Slice-100K dataset, capable of producing large volumes of layer-wise semantic segmentation masks. Through analysis of this large-scale synthetic data, we identify a systemic foreground–background class imbalance (1:24 ratio) inherent to AM monitoring, which causes standard Dice loss formulations to diverge catastrophically into a phenomenon we formalize as the “Dice Crash.” To effectively leverage large amounts of unlabeled data, we adapt the Meta Pseudo-Labeling (MPL) framework for industrial segmentation. We evaluate MPL’s true marginal utility by integrating it with both a standard U-Net and a robust state-of-the-art nnU-Net architecture. Experimental outputs show that while MPL yields substantial performance gains (+15.2%) on weak baselines, integrating it with an optimally configured strong baseline consistently improves segmentation accuracy and suppresses false foreground detections, thereby mitigating confirmation bias. These findings demonstrate that semi-supervised learning via continuous bilevel optimization offers a practical and robust enhancement to data-scarce additive manufacturing monitoring. Because any hidden defects in the topmost layer will be permanently buried by subsequent extrusion, this foundational layer-wise segmentation step is the most critical primitive of the monitoring pipeline.
- Open Access
- 26 December 2025
- Published by MDPI AG in Automation
- Vol. 7 (1) , 5
Reinforcement Learning for Uplink Access Optimization in UAV-Assisted 5G Networks Under Emergency Response
We study UAV-assisted 5G uplink connectivity for disaster response, in which a UAV (unmanned aerial vehicle) acts as an aerial base station to restore service to ground users. We formulate a joint control problem coupling UAV kinematics (bounded acceleration and velocity), per-subchannel uplink power allocation, and uplink non-orthogonal multiple access (UL-NOMA) scheduling with adaptive successive interference cancellation (SIC) under a minimum user-rate constraint. The wireless channel follows 3GPP urban macro (UMa) with probabilistic line of sight/non-line of sight (LoS/NLoS), realistic receiver noise levels and noise figure, and user equipment (UE) transmit-power limits. We propose a bounded-action proximal policy optimization with generalized advantage estimation (PPO-GAE) agent that parameterizes acceleration and power with squashed distributions and enforces feasibility by design. Across four user distributions (clustered, uniform, ring, and edge-heavy) and multiple rate thresholds, our method increases the fraction of users meeting the target rate by 8.2–10.1 percentage points compared to strong baselines (OFDMA with heuristic placement, PSO-based placement/power, and PPO without NOMA) while reducing median UE transmit power by . The results are averaged over at least five random seeds, with confidence intervals. Ablations isolate the gains from NOMA, adaptive SIC order, and bounded-action parameterization. We discuss robustness to imperfect SIC and CSI errors and release code/configurations to support reproducibility.- 26 December 2025
- Published by MDPI AG in Automation
- Vol. 7 (1) , 5
Open AccessReinforcement Learning for Uplink Access Optimization in UAV-Assisted 5G Networks Under Emergency Response
We study UAV-assisted 5G uplink connectivity for disaster response, in which a UAV (unmanned aerial vehicle) acts as an aerial base station to restore service to ground users. We formulate a joint control problem coupling UAV kinematics (bounded acceleration and velocity), per-subchannel uplink power allocation, and uplink non-orthogonal multiple access (UL-NOMA) scheduling with adaptive successive interference cancellation (SIC) under a minimum user-rate constraint. The wireless channel follows 3GPP urban macro (UMa) with probabilistic line of sight/non-line of sight (LoS/NLoS), realistic receiver noise levels and noise figure, and user equipment (UE) transmit-power limits. We propose a bounded-action proximal policy optimization with generalized advantage estimation (PPO-GAE) agent that parameterizes acceleration and power with squashed distributions and enforces feasibility by design. Across four user distributions (clustered, uniform, ring, and edge-heavy) and multiple rate thresholds, our method increases the fraction of users meeting the target rate by 8.2–10.1 percentage points compared to strong baselines (OFDMA with heuristic placement, PSO-based placement/power, and PPO without NOMA) while reducing median UE transmit power by . The results are averaged over at least five random seeds, with confidence intervals. Ablations isolate the gains from NOMA, adaptive SIC order, and bounded-action parameterization. We discuss robustness to imperfect SIC and CSI errors and release code/configurations to support reproducibility. - Open Access
- 24 September 2025
- Published by MDPI AG in Preprints.org
Reinforcement Learning for Uplink Access Optimization in UAV-Assisted 5G Networks Under Emergency Response
We study Unmanned Aerial Vehicle (UAV) assisted 5G uplink connectivity for disaster response, where a UAV acts as an aerial base station to restore service to ground users. We formulate a joint control problem coupling UAV kinematics (bounded acceleration and velocity), per-subchannel uplink power allocation, and uplink Non-Orthogonal Multiple Access (UL-NOMA) scheduling with adaptive successive interference cancellation (SIC) under a minimum user-rate constraint. The wireless channel follows 3GPP urban macro (UMa) with probabilistic Line of Sight/Non Line of Sight (LoS/NLoS), realistic receiver noise and noise figure, and user equipment (UE) transmit-power limits. We propose a bounded-action proximal policy optimization with generalized advantage estimation (PPO-GAE) agent that parameterizes acceleration and power with squashed distributions and enforces feasibility by design. Across four user distributions (clustered, uniform, ring, edge-heavy) and multiple rate thresholds, our method increases the fraction of users meeting the target rate by 8.2−10.1 percentage points over strong baselines (OFDMA with heuristic placement, PSO-based placement/power, PPO without NOMA) while reducing median UE transmit power by 64.6%. Results are averaged over ≥5 random seeds with 95% confidence intervals; ablations isolate the gains from NOMA, adaptive SIC order, and bounded action parameterization. We discuss robustness to imperfect SIC and CSI errors, and release code/configurations to support reproducibility.
- 24 September 2025
- Published by MDPI AG in Preprints.org
Open AccessReinforcement Learning for Uplink Access Optimization in UAV-Assisted 5G Networks Under Emergency Response
We study Unmanned Aerial Vehicle (UAV) assisted 5G uplink connectivity for disaster response, where a UAV acts as an aerial base station to restore service to ground users. We formulate a joint control problem coupling UAV kinematics (bounded acceleration and velocity), per-subchannel uplink power allocation, and uplink Non-Orthogonal Multiple Access (UL-NOMA) scheduling with adaptive successive interference cancellation (SIC) under a minimum user-rate constraint. The wireless channel follows 3GPP urban macro (UMa) with probabilistic Line of Sight/Non Line of Sight (LoS/NLoS), realistic receiver noise and noise figure, and user equipment (UE) transmit-power limits. We propose a bounded-action proximal policy optimization with generalized advantage estimation (PPO-GAE) agent that parameterizes acceleration and power with squashed distributions and enforces feasibility by design. Across four user distributions (clustered, uniform, ring, edge-heavy) and multiple rate thresholds, our method increases the fraction of users meeting the target rate by 8.2−10.1 percentage points over strong baselines (OFDMA with heuristic placement, PSO-based placement/power, PPO without NOMA) while reducing median UE transmit power by 64.6%. Results are averaged over ≥5 random seeds with 95% confidence intervals; ablations isolate the gains from NOMA, adaptive SIC order, and bounded action parameterization. We discuss robustness to imperfect SIC and CSI errors, and release code/configurations to support reproducibility.
- 6 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 01075-01084
Enhancing Power Quality in Smart Grids: Coordinated Control of Voltage and Current Harmonics Using Dual-Interface Converters
Harmonic distortion, caused by the superposition of multiples of the fundamental frequency, represents a significant source of electrical noise, degrading both system stability and power quality. This distortion can exacerbate supply voltage harmonics, especially when a single Distributed Generation (DG) interface converter is used to compensate for local load harmonic currents. The issue becomes particularly challenging when the main grid voltage is severely distorted. This paper presents an innovative approach to reducing current harmonics in a distorted distribution system by employing a dual-interface converter configuration. The proposed method enhances power quality by compensating for both supply voltage and grid current harmonics simultaneously. Simulation results demonstrate the effectiveness of this technique in mitigating harmonic distortion, improving system efficiency, and maintaining the Total Harmonic Distortion (THD) of the grid voltage within permissible limits. The results underscore the potential of dual-interface converters to provide superior harmonic compensation, making them a viable solution for modern smart grid systems.
- 6 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 01075-01084
Enhancing Power Quality in Smart Grids: Coordinated Control of Voltage and Current Harmonics Using Dual-Interface Converters
Harmonic distortion, caused by the superposition of multiples of the fundamental frequency, represents a significant source of electrical noise, degrading both system stability and power quality. This distortion can exacerbate supply voltage harmonics, especially when a single Distributed Generation (DG) interface converter is used to compensate for local load harmonic currents. The issue becomes particularly challenging when the main grid voltage is severely distorted. This paper presents an innovative approach to reducing current harmonics in a distorted distribution system by employing a dual-interface converter configuration. The proposed method enhances power quality by compensating for both supply voltage and grid current harmonics simultaneously. Simulation results demonstrate the effectiveness of this technique in mitigating harmonic distortion, improving system efficiency, and maintaining the Total Harmonic Distortion (THD) of the grid voltage within permissible limits. The results underscore the potential of dual-interface converters to provide superior harmonic compensation, making them a viable solution for modern smart grid systems.
- 9 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2919-2924
Unveiling the Future: A Comparative Analysis of LSTM and SP-LSTM for Network Traffic Prediction in 6G Networks
The emergence of 6G connectivity heralds a transformative era in wireless communication, emphasizing the necessity of network-wide intelligence for fully automated operations. At the heart of this paradigm shift lies the crucial need for an efficient algorithm capable of accurately predicting network traffic dynamics. This paper presents a comprehensive comparative study between two pivotal neural network architectures, LSTM (Long Short-Term Memory), and SP-LSTM (Speed-Optimized LSTM), within the context of network traffic prediction for the burgeoning 6G landscape. While LSTM stands as a widely acknowledged recurrent neural network, our innovative SP-LSTM is meticulously engineered for rapid and efficient decision-making. Through rigorous evaluation in a controlled environment, both models are examined for their efficacy in forecasting network traffic patterns. LSTM's proficiency in capturing long-term dependencies in sequential data is juxtaposed against SP-LSTM's emphasis on delivering swift and precise predictions. Our comparative analysis elucidates the distinctive strengths of these algorithms, assessing LSTM's effectiveness under varying conditions and scrutinizing SP-LSTM's proficiency in providing rapid, accurate forecasts. Additionally, we provide a GitHub link for accessing the project. This study offers vital insights into the relative merits and limitations of LSTM and SP-LSTM in network traffic prediction, essential for the advancement of intelligent 6G networking. These insights inform the selection of the optimal algorithm for realizing a seamless, intelligent 6G future with unparalleled capabilities.
- 9 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2919-2924
Unveiling the Future: A Comparative Analysis of LSTM and SP-LSTM for Network Traffic Prediction in 6G Networks
The emergence of 6G connectivity heralds a transformative era in wireless communication, emphasizing the necessity of network-wide intelligence for fully automated operations. At the heart of this paradigm shift lies the crucial need for an efficient algorithm capable of accurately predicting network traffic dynamics. This paper presents a comprehensive comparative study between two pivotal neural network architectures, LSTM (Long Short-Term Memory), and SP-LSTM (Speed-Optimized LSTM), within the context of network traffic prediction for the burgeoning 6G landscape. While LSTM stands as a widely acknowledged recurrent neural network, our innovative SP-LSTM is meticulously engineered for rapid and efficient decision-making. Through rigorous evaluation in a controlled environment, both models are examined for their efficacy in forecasting network traffic patterns. LSTM's proficiency in capturing long-term dependencies in sequential data is juxtaposed against SP-LSTM's emphasis on delivering swift and precise predictions. Our comparative analysis elucidates the distinctive strengths of these algorithms, assessing LSTM's effectiveness under varying conditions and scrutinizing SP-LSTM's proficiency in providing rapid, accurate forecasts. Additionally, we provide a GitHub link for accessing the project. This study offers vital insights into the relative merits and limitations of LSTM and SP-LSTM in network traffic prediction, essential for the advancement of intelligent 6G networking. These insights inform the selection of the optimal algorithm for realizing a seamless, intelligent 6G future with unparalleled capabilities.
- 29 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 38-44
Reinforcement Learning-Driven Adaptive Traffic Routing Role Delegation for Enhanced SDN Network Performance
In this paper, we introduce a novel method for autonomously assigning routing traffic roles in Software Defined Networks (SDN) using reinforcement learning. Our approach centers around a cloud-hosted central SDN controller interacting with two sub-SDN controllers situated in edge data centers. Each sub-controller manages three layer 2 switches and one wireless access point, connecting to five wired or wireless hosts generating randomized traffic. By collecting network data from hostto-host communications across diverse applications, we train a reinforcement learning agent to facilitate adaptive routing role delegation. The agent is tailored to identify optimal links between the main controller and subcontrollers, minimizing traffic congestion and ensuring ultra-low latency and high reliability during traffic routing. Our research aims to achieve dynamic and adaptive optimal routing, accommodating fluctuating traffic patterns and new device additions to the network. Leveraging PyTorch, we implement a neural network for the agent and employ a Q-learning algorithm for training, optimizing a reward function based on observed traffic congestion. Once trained, the agent autonomously delegates routing traffic roles to the least congested sub-controller. Performance evaluation involves analyzing traffic load and latency metrics, demonstrating the agent’s proficiency in identifying optimal links and enhancing network performance, as observed through extensive experimentation and evaluation across diverse SDN scenarios. Furthermore, to foster collaboration and reproducibility within the research community, we make our implementation code and dataset publicly available on GitHub at https://github.com/your_username/project_name.
- 29 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 38-44
Reinforcement Learning-Driven Adaptive Traffic Routing Role Delegation for Enhanced SDN Network Performance
In this paper, we introduce a novel method for autonomously assigning routing traffic roles in Software Defined Networks (SDN) using reinforcement learning. Our approach centers around a cloud-hosted central SDN controller interacting with two sub-SDN controllers situated in edge data centers. Each sub-controller manages three layer 2 switches and one wireless access point, connecting to five wired or wireless hosts generating randomized traffic. By collecting network data from hostto-host communications across diverse applications, we train a reinforcement learning agent to facilitate adaptive routing role delegation. The agent is tailored to identify optimal links between the main controller and subcontrollers, minimizing traffic congestion and ensuring ultra-low latency and high reliability during traffic routing. Our research aims to achieve dynamic and adaptive optimal routing, accommodating fluctuating traffic patterns and new device additions to the network. Leveraging PyTorch, we implement a neural network for the agent and employ a Q-learning algorithm for training, optimizing a reward function based on observed traffic congestion. Once trained, the agent autonomously delegates routing traffic roles to the least congested sub-controller. Performance evaluation involves analyzing traffic load and latency metrics, demonstrating the agent’s proficiency in identifying optimal links and enhancing network performance, as observed through extensive experimentation and evaluation across diverse SDN scenarios. Furthermore, to foster collaboration and reproducibility within the research community, we make our implementation code and dataset publicly available on GitHub at https://github.com/your_username/project_name.
- 29 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 0061-0067
Smart Meter Performance Under Wired and Wireless Cyber Security Attack
Utility companies have deployed millions of smart meters to remotely gather real-time data on power consumption from client premises on a regular basis. It provides utility service providers with the power consumption data to remotely manage their customers’ electricity usage. Given the importance of these data, it’s crucial to make sure that any information collected locally at smart meters or transmitted remotely via wired or wireless network is trustworthy, private and secure. The security risks associated with smart meters are constantly evolving, and more security breaches occurring frequently. Research on data security breaches for wired or wireless smart meters—whether available for residential or commercial use and how they affect the finances of the companies that deploy them is currently lacking. In this work, we conducted experiments to assess the performance of wired and wireless smart meters under cyber security attacks, with the goal of comparing their performance under different network configurations, and to understand the financial implications for utility companies for the compromised power consumption data.
- 29 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 0061-0067
Smart Meter Performance Under Wired and Wireless Cyber Security Attack
Utility companies have deployed millions of smart meters to remotely gather real-time data on power consumption from client premises on a regular basis. It provides utility service providers with the power consumption data to remotely manage their customers’ electricity usage. Given the importance of these data, it’s crucial to make sure that any information collected locally at smart meters or transmitted remotely via wired or wireless network is trustworthy, private and secure. The security risks associated with smart meters are constantly evolving, and more security breaches occurring frequently. Research on data security breaches for wired or wireless smart meters—whether available for residential or commercial use and how they affect the finances of the companies that deploy them is currently lacking. In this work, we conducted experiments to assess the performance of wired and wireless smart meters under cyber security attacks, with the goal of comparing their performance under different network configurations, and to understand the financial implications for utility companies for the compromised power consumption data.