A Quantum Genetic Algorithm-Enhanced Self-Supervised Intrusion Detection System for Wireless Sensor Networks in the Internet of Things

Hamid Barati

REVIEW 4 major objections 4 minor 28 references

A Quantum Genetic Algorithm-Enhanced Self-Supervised Intrusion Detection System for Wireless Sensor Networks in the Internet of Things

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A hybrid quantum-genetic and self-supervised intrusion detection system is reported to reach 96.7% accuracy on NSL-KDD and 95.2% on UNSW-NB15 while running on edge-class hardware in under 300 MB of memory.

desk verdict A coherent but unproven hybrid; the QGA+SSL pipeline is a reasonable extension, but the reported gains and resource numbers are unsupported by controlled experiments. read the letter →

arxiv 2509.03744 v1 pith:BR5LPAQ6 submitted 2025-09-03 cs.CR

Hamid Barati This is my paper · ORCID

classification cs.CR

keywords intrusiondetectionwirelesssensornetworksInternetofThingsquantumgeneticalgorithmself-supervisedlearningcontrastivefeatureselectionedgedeployment

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that intrusion detection for wireless sensor networks and IoT devices can be made both accurate and lightweight by combining self-supervised learning, which trains on unlabeled traffic, with a quantum-inspired genetic algorithm that selects features and tunes the classifier. The proposed QGA-SSL IDS is reported to reach 96.7% accuracy with a 3.2% false positive rate on NSL-KDD, and 95.2% accuracy with a 4.1% false positive rate on UNSW-NB15, beating the FS3, WOGRU-IDS, and Lightweight IDS baselines on all detection metrics. The same system is said to run in under 300 MB of memory at roughly 1,500 packets per second on Raspberry Pi-class devices, which is what would make it practical for real WSN and IoT deployments.

What carries the argument

The load-bearing mechanism is the pairing of a self-supervised encoder with a quantum-inspired evolutionary search. The SSL encoder maps augmented views of each traffic instance to compact embeddings through a contrastive loss, giving the classifier a label-free representation of normal and malicious traffic. QGA then represents candidate feature subsets and hyperparameters as qubit chromosomes, each qubit a pair of probability amplitudes, and evolves them with quantum rotation gates, repeatedly measuring the chromosomes into concrete configurations and scoring them with a fitness function that combines accuracy, false positive rate, and cost. This quantum-inspired search, not actual quantum

What would settle it

Run all four methods (QGA-SSL, FS3, WOGRU-IDS, Lightweight IDS) on identical train/test splits with identical preprocessing and class weighting, and also run QGA-SSL with the QGA stage replaced by random feature selection. If the baselines close the gap to within one accuracy point, or if random selection performs as well as QGA, the paper's central attribution collapses.

Watch

Extended reading notes

Core claim

The central claim is that one pipeline can address the two central problems of IoT intrusion detection: scarce labeled data and scarce compute. The paper's IDS first trains a self-supervised encoder on unlabeled traffic using contrastive and auxiliary pretext tasks, then applies a quantum genetic algorithm to evolve a feature subset and classifier parameters, with a fitness function that balances accuracy, false positive rate, and computational cost. On NSL-KDD the full system reports 96.7% accuracy and 3.2% FPR; on UNSW-NB15 it reports 95.2% accuracy and 4.1% FPR, outperforming FS3, WOGRU-IDS, and Lightweight IDS while taking less training time than all but the Lightweight baseline. Deploym

Load-bearing premise

The paper's claimed advantage rests on the assumption that its higher accuracy and lower false positive rate come from QGA feature selection and SSL representation learning, not from differences in how the datasets were split, preprocessed, weighted, or which classifier was used; no ablation or controlled re-implementation of the baselines is reported to rule those out.

Editorial extensions

If this is right

If the reported results hold, IDS training no longer depends on large labeled attack corpora; SSL pretraining consumes unlabeled traffic that WSNs generate anyway.
A low false positive rate (3.2% and 4.1%) matters operationally: in energy-limited sensor networks, false alarms waste battery and operator attention.
Sub-300 MB memory and roughly 1,500 packets per second place the model in range of Raspberry Pi-class nodes, not just cloud servers.
The combined pipeline has polynomial complexity O(N L m^2 + T P N k), in contrast to transformer-based IDS scaling as O(N L d^2), supporting edge deployment.
The fitness function's cost term means the same architecture can be re-optimized for different hardware budgets without changing the model family.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: the paper never ablates the two components, so a natural next experiment is to train the SSL encoder alone, the classifier with QGA alone, and the full stack on identical splits; that would show whether the accuracy gain comes from QGA, from SSL, or from their interaction.
Editorial inference: because the baselines are re-used from their published numbers, the headline margins could shrink under a common preprocessing and class-weighting protocol; re-running all four methods on one standardized split would be the cleanest check.
Editorial inference: QGA's fitness function already includes a cost term, so one could extend the paper by optimizing directly for energy consumption per packet rather than memory as a proxy, a more direct metric for battery-powered WSN nodes.
Editorial inference: the SSL component is evaluated only on two datasets; testing on a newer traffic corpus such as CIC-IDS2017 or CICIoT2023 would show whether the learned representations transfer beyond NSL-KDD and UNSW-NB15.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper combines QGA with SSL for intrusion detection, which is a legitimate extension of two existing lines of work rather than a new mechanism. The architecture is clearly described: an SSL encoder learns embeddings, QGA selects features and tunes the classifier, and a multi-objective fitness balances accuracy, FPR, and cost. That integration is sensible, and the complexity analysis, while coarse, at least gestures at deployment constraints. If the numbers held up, this would be a modest, useful step for resource-limited IoT settings.

The problem is that the experimental evidence does not support the claims. Tables 1 and 2 report single-point accuracy and FPR with no error bars or repeated runs. Baselines are quoted from other papers rather than re-run under identical splits and preprocessing, so the claimed improvements could come from evaluation protocol rather than the method. There is no ablation to separate QGA from SSL contributions, and the paper omits the encoder architecture, hyperparameters, and dataset split details. The resource claims are also unmeasured and internally inconsistent (sub-300 MB in Section 3.5, less than 220 MB in Section 4.4), with no procedure or raw data given.

I do not see a circularity problem in using accuracy and FPR inside the fitness function and later reporting them; that is standard model selection. But the absence of an independent test set and controlled baselines makes the performance numbers unverifiable. The self-citation density is higher than needed, but it does not affect the technical argument.

Who is this for? A newcomer to SSL or QGA-based IDS might get a concise walkthrough of the components, but anyone working in the area will need stronger evidence before trusting the results. In its current form, the load-bearing claims are unsupported, and I would not send it to peer review. A revised version that re-runs baselines under identical conditions, adds ablations, reports variance, and measures actual resource usage could deserve referee time. As it stands, it is a desk reject with an invitation to resubmit after substantial revision.

Referee Report

4 major / 4 minor

Summary. The manuscript proposes a hybrid intrusion detection system for IoT/WSN environments that combines Self-Supervised Learning (SSL) for representation learning with a Quantum Genetic Algorithm (QGA) for feature selection and classifier tuning. The authors report accuracy/FPR figures on NSL-KDD and UNSW-NB15, claim superiority over three baselines, and assert deployment feasibility on Raspberry Pi-class devices. The paper contains a high-level pipeline description, loss functions for contrastive SSL, a QGA fitness function, a complexity analysis, and two comparison tables. The central empirical claims are that the proposed QGA-SSL IDS achieves 96.7% accuracy / 3.2% FPR on NSL-KDD and 95.2% / 4.1% on UNSW-NB15, outperforming FS3, WOGRU-IDS, and Lightweight IDS, and that it runs within 220–300 MB memory at about 1,500 packets/s on a Raspberry Pi.

Significance. If the proposed method were properly validated, the combination of SSL with quantum-inspired evolutionary optimization would be a plausible contribution to resource-constrained IoT security. The manuscript is clearly structured and the idea is coherent. However, the current evidence is not sufficient to establish the claimed advantages. The experiments lack controlled reimplementation, ablations, error bars, and enough implementation detail for reproducibility. The resource claims are unmeasured. The paper also contains citation mismatches that make the baselines impossible to identify. The contribution is therefore, at this stage, only a promising concept rather than a demonstrated result.

major comments (4)

[Sec. 4.3, Tables 1–2] The tables report single-point accuracies, FPRs, and training times with no error bars, repeated runs, or statistical tests. More seriously, the baseline numbers are quoted from references [3], [10], and [13], but in the reference list those citations are not the FS3, WOGRU-IDS, and Lightweight IDS papers; they are unrelated WSN/key-management and fuzzy-detection works. The comparisons therefore cannot be verified and may not correspond to the named methods. Identical preprocessing, splits, and evaluation protocols are not guaranteed, so the reported gains cannot be attributed to the proposed method.
[Sec. 4.1; Sec. 3.2–3.4] The experimental setup omits the information needed to reproduce the results: the SSL encoder architecture (depth, width, augmentation scheme), the projection head, the downstream classifier C_phi, the QGA population size, number of generations, rotation-angle schedule, all hyperparameters (tau, lambda_c, lambda_m, lambda_t), and the trade-off weights alpha, beta, gamma. Dataset split ratios, preprocessing details, and class-imbalance handling are also not specified. Without these, the reported numerical results are not reproducible.
[Sec. 3.5 and Sec. 4.4] The resource-usage claim is internally inconsistent: Section 3.5 states 'sub-300 MB memory', while Section 4.4 says 'less than 220 MB'. Both claims are presented as 'Empirical tests confirm' or 'Testing on Raspberry Pi confirmed', but no measurement methodology, no raw data, no variance, and no details of the Raspberry Pi setup are given. The throughput figure of ~1,500 packets/s is similarly unsupported.
[Sec. 3.4 and Sec. 4.3] There is no ablation isolating the contributions of QGA and SSL. The claim that the method's advantage comes from the 'synergy' of the two components is not supported. In addition, the manuscript does not state whether the QGA fitness (which includes accuracy and FPR) is evaluated on a validation set or on the test set; if fitness were computed on the test set, the reported results would be optimistically biased. This must be clarified.

minor comments (4)

[Throughout] The paper uses 'Fig. X' as a placeholder instead of an actual figure number, and Section 4.4 contains multiple standalone paragraphs of discussion that would benefit from tighter organization.
[Sec. 3.2] The NT-Xent loss equation is missing a parenthesis in the denominator; the mathematical notation should be corrected for clarity.
[References] The reference list is inconsistently formatted: the first two entries have no numbers, while the list restarts at '1.' mid-page. The in-text baseline citations [3], [10], [13] do not match the intended FS3/WOGRU/Lightweight references, which appear elsewhere in the related-work table as [16], [23], and [26].
[Sec. 4.1] The evaluation metrics list includes Training Time, but the resource-constrained feasibility discussion focuses on memory and throughput; the relationship between reported training time and deployment claims should be clarified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; QGA fitness uses accuracy/FPR but reported results are empirical, not definitional.

full rationale

I examined the claimed derivation chain. The only apparent candidate for circularity is that the QGA fitness function F(S)=α·Acc(S)+β·(1−FPR(S))−γ·Cost(S) includes the same metrics (accuracy, FPR) that are later reported in Tables 1-2. This is not a circular derivation: it is a standard model-selection loop. The reported values are empirical outcomes of the selected model M*(x)=C_phi*(E_theta(x~)_S*) evaluated on benchmark datasets, not values computed from the fitness definition. No equation in the paper defines the reported accuracy or FPR as equal to the fitness function, and the paper gives no uniqueness theorem or imported ansatz that forces the result. The baseline numbers in Tables 1-2 are quoted from external papers and are not re-run under identical protocols, but that is an uncontrolled-comparison/evidence weakness, not a self-referential reduction. The numerous self-citations (references 2-13) are background material on WSN key management, coverage, and routing; none is used to justify the central QGA-SSL claim, so they do not create load-bearing circularity. The absence of ablations and omitted SSL/QGA hyperparameters are completeness/reproducibility gaps, not circularity. Consequently no circular step can be quoted and reduced by construction; the finding is no significant circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method relies on many undisclosed hyperparameters and on assumptions that benchmark datasets and quoted baseline results represent the WSN/IoT deployment scenario. No independent evidence is provided for the SSL embedding quality or the QGA search advantage.

free parameters (5)

Fitness trade-off weights (alpha, beta, gamma)
The fitness function F(S) = alpha*Acc + beta*(1-FPR) - gamma*Cost uses three weights to balance accuracy, FPR, and computational cost; values are never specified.
SSL loss weights (lambda_c, lambda_m, lambda_t)
The joint SSL objective weighs contrastive, masking, and temporal losses with lambdas never specified.
Temperature tau in contrastive loss
The contrastive loss uses a temperature hyperparameter not disclosed.
QGA rotation angle delta_theta
The quantum rotation gate update rule and angle adaptation are not specified.
Classifier hyperparameters
The lightweight classifier C_phi and its hyperparameters are not described.

assumptions (4)

domain assumption NSL-KDD and UNSW-NB15 are representative of wireless sensor network IoT traffic.
Section 4.1 uses these two benchmark datasets to evaluate a system designed for WSN/IoT, but these are generic network intrusion datasets without sensor-specific characteristics.
domain assumption The self-supervised pretext tasks (contrastive, temporal prediction, feature masking) learn representations useful for intrusion detection.
Section 3.2 assumes these tasks yield discriminative embeddings, but no analysis of embedding quality is provided.
domain assumption Quoted baseline results from [3], [10], and [13] are accurate and directly comparable.
Section 4.3 compares against numbers taken from prior papers rather than re-implementing baselines under identical conditions.
standard math The standard quantum-inspired qubit and rotation gate formalism is assumed without needing external verification.
Section 3.3 uses the textbook QGA formulation, which is unproblematic.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Quantum Genetic Algorithm-Enhanced Self-Supervised Intrusion Detection System for Wireless Sensor Networks in the Internet of Things." pith.science (2026). https://pith.science/paper/BR5LPAQ6

@misc{pith2026250903744,
  author       = {Pith},
  title        = {Pith review of: A Quantum Genetic Algorithm-Enhanced Self-Supervised Intrusion Detection System for Wireless Sensor Networks in the Internet of Things},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BR5LPAQ6}},
  note         = {Machine review of arXiv:2509.03744}
}

read the original abstract

The rapid expansion of the Internet of Things (IoT) and Wireless Sensor Networks (WSNs) has significantly increased the attack surface of such systems, making them vulnerable to a wide range of cyber threats. Traditional Intrusion Detection Systems (IDS) often fail to meet the stringent requirements of resource-constrained IoT environments due to their high computational cost and reliance on large labeled datasets. To address these challenges, this paper proposes a novel hybrid Intrusion Detection System that integrates a Quantum Genetic Algorithm (QGA) with Self-Supervised Learning (SSL). The QGA leverages quantum-inspired evolutionary operators to optimize feature selection and fine-tune model parameters, ensuring lightweight yet efficient detection in resource-limited networks. Meanwhile, SSL enables the system to learn robust representations from unlabeled data, thereby reducing dependency on manually labeled training sets. The proposed framework is evaluated on benchmark IoT intrusion datasets, demonstrating superior performance in terms of detection accuracy, false positive rate, and computational efficiency compared to conventional evolutionary and deep learning-based IDS models. The results highlight the potential of combining quantum-inspired optimization with self-supervised paradigms to design next-generation intrusion detection solutions for IoT and WSN environments.

Figures

Figures reproduced from arXiv: 2509.03744 by the authors.

**Figure 1.** System Architecture of QGA-SSL IDS Data Collection Preprocessing SSL Encoder QGA Optimization Intrusion Detection Response [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 23 canonical work pages

[3]

Barati, A., Dehghan, M., Barati, H., & Mazreah, A. A. (2008, August). Key management mechanisms in wireless sensor networks. In 2008 Second International Conference on Sensor Technologies and Applications (sensorcomm 2008) (pp. 81-86). IEEE

work page 2008
[10]

Barati, H., Movaghar, A., Barati, A., & Azizi Mazresh, A. (2008). A review of coverage and routing for wireless sensor networks. World Acad. Sci. Eng. Tech, 37, 296-302

work page 2008
[13]

J., Movaghar, A., & At tarzadeh, I

Barati, A., Dastgheib, S. J., Movaghar, A., & At tarzadeh, I. (2012). An effective fuzzy based algorithm to detect faulty readings in long thin wireless sensor networks. International Journal on Technical and Physical Problems of Engineering (IJTPE), 3(1), 2077-3528

work page 2012
[1]

Atzori, L., Iera, A., & Morabito, G. (2010). The internet of things: A survey. Computer networks, 54(15), 2787-2805

work page 2010
[2]

Hallafi, A., Barati, A., & Barati, H. (2023). A distributed energy -efficient coverage holes detection and recovery method in wireless sensor networks using the grasshopper optimization algorithm. Journal of Ambient Intelligence and Humanized Computing, 14(10), 13697-13711

work page 2023
[4]

A., Barati, A., & Barati, H

Khah, S. A., Barati, A., & Barati, H. (2023). A dynamic and multi-level key management method in wireless sensor networks (WSNs). Computer Networks, 236, 109997

2023
[5]

Barati, A., Movaghar, A., & Sabaei, M. (2016). RDTP: Reliable data transport protocol in wireless sensor networks. Telecommunication Systems, 62(3), 611-623

2016
[6]

Da Xu, L., He, W., & Li, S. (2014). Internet of things in industries: A survey. IEEE Transactions on industrial informatics, 10(4), 2233-2243

work page 2014

Show all 28 references

[7]

Barati, A., Movaghar, A., & Sabaei, M. (2014). Energy efficient and high speed error control scheme for real time wireless sensor networks. International Journal of Distributed Sensor Networks, 10(5), 698125

2014
[8]

J., & Attarza deh, I

Javanmardi, S., Barati, A., Dastgheib, S. J., & Attarza deh, I. (2012). A novel approach for faulty node detection with the aid of fuzzy theory and majority voting in wireless sensor networks. International Journal of Advanced Smart Sensor Network Systems, 2(4), 1-10

2012
[9]

Barati, A., Movaghar, A., Modiri, S., & Sabaei, M. (2012). A reliable & energy-efficient scheme for real time wireless sensor networks applications. Basic. Appl. Sci. Res, 2(10), 10150-10157

2012
[11]

Shojarazavi, T., Barati, H., & Barati, A. (2022). A wrapper method based on a modified two -step league championship algorithm for detecting botnets in IoT environments. Computing, 104(8), 1753-1774

2022
[12]

Bozorgzadeh, E., Barati, H., & Barati, A. (2020). 3DEOR: an opportunity routing protocol using evidence theory appropriate for 3D urban environments in VANETs. IET Communications, 14(22), 4022-4028

2020
[14]

Xu, R., Wu, G., Wang, W., Gao, X., He , A., & Zhang, Z. (2024). Applying self -supervised learning to network intrusion detection for network flows with graph neural network. Computer Networks, 248, 110495

2024
[15]

(2024, December)

Tan, X., Cheng, J., Li, H., & Yang, Y. (2024, December). Contrastive Learning for Netwo rk Intrusion Detection: A Comprehensive Survey. In Proceedings of the 2024 2nd International Conference on Computer, Internet of Things and Smart City (pp. 160-166)

2024
[16]

(2023, December)

Ayesha S, D., & AB, S. (2023, December). FS3: Few -shot and self -supervised framework for efficient intrusion detection in Internet of Things networks. In Proceedings of the 39th Annual Computer Security Applications Conference (pp. 138-149)

2023
[17]

Li, L., Lu, Y., Yang, G., & Yan, X. (2024). End -to-end network intrusion detection based on contrastiv e learning. Sensors, 24(7), 2122

2024
[18]

Koukoulis, I., Syrigos, I., & Korakis, T. (2025). Self -Supervised Transformer-based Contrastive Learning for Intrusion Detection Systems. arXiv preprint arXiv:2505.08816

2025 arXiv
[19]

(2024, November)

Ben Atitallah, S., Driss, M., Boulila, W., & Koub aa, A. (2024, November). Enhancing Internet of Things Security Through Self-supervised Graph Neural Networks. In International Conference on Smart Systems and Emerging Technologies (pp. 186-197). Cham: Springer Nature Switzerland

2024
[20]

Liu, L., & Xu, M. (2025). A network intrusion detection method based on contrastive learning and Bayesian Gaussian Mixture Model. Cybersecurity, 8(1), 59

2025
[21]

Albanbay, N., Tursynbek, Y., Graffi, K., Uskenbayeva, R., Kalpeyeva, Z., Abilkaiyr, Z., & Ayapov, Y. (2025). Federated Learning-Based Intrusion Detection in IoT Networks: Performance Evaluation and Data Scaling Study. Journal of Sensor and Actuator Networks, 14(4), 78

2025
[22]

Olanrewaju-George, B., & Pranggono, B. (2025). Federated learning -based intrusion detection system for the internet of things using unsupervised and supervised deep learning models. Cyber Security and Applications, 3, 100068

2025
[23]

H., Narayana, C

Ramana, K., Revathi, A., Gayathri, A., Jhaveri, R. H., Narayana, C. L., & Kumar, B. N. (2022). WOGRU - IDS—An intelligent intrusion de tection system for IoT assisted Wireless Sensor Networks. Computer Communications, 196, 195-206

2022
[24]

S., Khan, M

Farhan, M., Waheed Ud Din, H., Ullah, S., Hussain, M. S., Khan, M. A., Mazhar, T., ... & Jaghdam, I. H. (2025). Network-based intrusion detection using deep learning technique. Scientific Reports, 15(1), 25550

2025
[25]

Yu, Y., Fu, Y., Liu, T., Wang, K., & An, Y. (2025). An attack detection method based on deep learning for internet of things. Scientific Reports, 15(1), 28812

2025
[26]

Wang, Z., Chen, H., Yang, S., Luo, X., Li, D., & Wang, J. (2023). A lightweight intrusion detection method for IoT based on deep learning and dynamic quantization. PeerJ Computer Science, 9, e1569

2023
[27]

Ling, Z., Qi, G., & Min, H. (2024). Intrusion detection using rough -fuzzy set and parallel quantum genetic algorithm. Journal of High Speed Networks, 30(1), 69-81

2024
[28]

Luqman, M., Zeeshan, M., Riaz, Q., Hussain, M., Tahir, H., Mazhar, N., & Khan, M. S. (2025). Intelligent parameter-based in-network IDS for IoT using UNSW-NB15 and BoT-IoT datasets. Journal of the Franklin Institute, 362(1), 107440

2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [3]

Barati, A., Dehghan, M., Barati, H., & Mazreah, A. A. (2008, August). Key management mechanisms in wireless sensor networks. In 2008 Second International Conference on Sensor Technologies and Applications (sensorcomm 2008) (pp. 81-86). IEEE

work page 2008

[2] [10]

Barati, H., Movaghar, A., Barati, A., & Azizi Mazresh, A. (2008). A review of coverage and routing for wireless sensor networks. World Acad. Sci. Eng. Tech, 37, 296-302

work page 2008

[3] [13]

J., Movaghar, A., & At tarzadeh, I

Barati, A., Dastgheib, S. J., Movaghar, A., & At tarzadeh, I. (2012). An effective fuzzy based algorithm to detect faulty readings in long thin wireless sensor networks. International Journal on Technical and Physical Problems of Engineering (IJTPE), 3(1), 2077-3528

work page 2012

[4] [1]

Atzori, L., Iera, A., & Morabito, G. (2010). The internet of things: A survey. Computer networks, 54(15), 2787-2805

work page 2010

[5] [2]

Hallafi, A., Barati, A., & Barati, H. (2023). A distributed energy -efficient coverage holes detection and recovery method in wireless sensor networks using the grasshopper optimization algorithm. Journal of Ambient Intelligence and Humanized Computing, 14(10), 13697-13711

work page 2023

[6] [4]

A., Barati, A., & Barati, H

Khah, S. A., Barati, A., & Barati, H. (2023). A dynamic and multi-level key management method in wireless sensor networks (WSNs). Computer Networks, 236, 109997

2023

[7] [5]

Barati, A., Movaghar, A., & Sabaei, M. (2016). RDTP: Reliable data transport protocol in wireless sensor networks. Telecommunication Systems, 62(3), 611-623

2016

[8] [6]

Da Xu, L., He, W., & Li, S. (2014). Internet of things in industries: A survey. IEEE Transactions on industrial informatics, 10(4), 2233-2243

work page 2014