Federated Learning and LLM-Driven Threat Intelligence for Zero Trust IoT Architecture

Amal Alshehri; Cihan Tunc

REVIEW 4 major objections 6 minor 15 references

Federated Learning and LLM-Driven Threat Intelligence for Zero Trust IoT Architecture

T0 review · 4 major / 6 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper claims that combining a federated autoencoder with a fine-tuned language model yields a privacy-preserving zero-trust architecture for IoT that detects every tested attack with perfect recall and automatically blocks the attacker.

desk verdict An honest engineering assembly whose headline numbers are artifacts of a circular evaluation; the architecture is worth reading as a prototype, not as a validated defense. read the letter →

arxiv 2607.17035 v1 pith:H22DNXET submitted 2026-07-19 cs.CR

Amal Alshehri , Cihan Tunc This is my paper · ORCID

classification cs.CR

keywords federatedlearningzerotrustarchitectureIoTsecurityanomalydetectionautoencoderLLMthreatintelligenceMQTT/TLSautomatedresponse

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 zero trust can be made practical for IoT by layering a federated autoencoder, a fine-tuned large language model, and automated firewall enforcement into a single pipeline. The autoencoder learns only normal traffic patterns on each device and flags anomalies without sharing raw data; the LLM converts each anomaly into a structured threat report with severity and remediation commands; and the broker automatically blocks malicious source IPs. The authors report perfect recall and a ROC-AUC of 1.0 across four attack types (port scan, DDoS SYN flood, TLS downgrade, plain MQTT) on a testbed of single-board computers and sensors extended with bootstrapped virtual devices. They interpret this as evidence that privacy-preserving anomaly detection and autonomous response can coexist on resource-constrained IoT hardware.

What carries the argument

The load-bearing mechanism is the combination of (1) a federated autoencoder that learns only normal traffic patterns and flags anything whose reconstruction error exceeds the p95 threshold; (2) FedAvg aggregation so raw sensor and network data never leave the devices; and (3) an LLM fine-tuned on 500 instruction–response pairs that converts anomaly feature vectors into structured JSON threat reports, which the system turns into iptables rules to block malicious IPs. The p95 threshold is the fulcrum: it sets a fixed 5% false-positive rate and, together with perfect reconstruction-error separation, yields the reported precision and recall.

What would settle it

Run the trained autoencoder and LLM pipeline on a live packet capture of a real port-scan or SYN-flood campaign against an MQTT broker and check whether recall stays at 1.0; or measure the fraction of virtual test samples that are near-duplicates of training samples (e.g., by Euclidean distance in feature space) and see whether AUC drops when those near-duplicates are removed.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a federated-learning autoencoder, aggregated with FedAvg over 25 clients and thresholded at the 95th percentile of reconstruction error, separates all tested attack traffic from normal MQTT/TLS traffic with zero false negatives, and that a LoRA-fine-tuned LLM maps the resulting anomaly features to accurate, actionable decisions (verdict, severity, remediation commands) that the broker can execute via firewall rules in under a second after the LLM responds. The authors see this as evidence that a zero-trust IoT architecture—mutual TLS on every MQTT connection, no implicit trust, continuous identity verification, and automated threat isol

Load-bearing premise

The whole evaluation hinges on the assumption that the synthetic attack traffic and bootstrapped virtual devices behave like real malicious IoT traffic; if that assumption fails, the perfect recall and AUC of 1.0 are artifacts of how the test data was manufactured.

Editorial extensions

If this is right

If the claim holds, zero trust for IoT does not require heavyweight per-device security: a broker-side monitor plus local autoencoders can enforce 'never trust, always verify' at the communication layer.
Privacy-preserving anomaly detection is feasible on low-power devices: clients share only weight updates, and per-sample inference latency is about 0.0065 ms.
Automated detection-to-response is achievable without human intervention: the LLM decides, and firewall enforcement completes in under a second; the main latency (about 41 seconds) comes from remote LLM inference.
The fixed 5% false-positive rate is a tunable design point: at high imbalance (up to 100:1 attack scarcity), recall stays perfect, so no attacks are missed, though precision drops.
A direct corollary is that the same pipeline generalizes across four distinct attack classes with a single shared feature set, suggesting the anomaly detector is attack-agnostic.

Reading between the lines

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

The reported ROC-AUC of 1.0 and perfect recall are only meaningful if the synthetic attack samples (built by modifying features of normal traffic) and the bootstrapped virtual devices are representative; if they share near-duplicates with the training distribution, the numbers could be optimistic.
A natural testable extension: replace the synthetic attacks with packet captures from real IoT botnets or published intrusion datasets and check whether recall and AUC survive.
The roughly 41-second LLM response delay, while stable, is larger than typical IoT alert latency budgets; a future design could cache responses, run a smaller local model, or limit LLM invocation to high-severity anomalies.
Because the autoencoder is trained only on normal traffic, the approach might detect novel (zero-day) attacks as anomalies, but the LLM's ability to classify them depends on how similar they are to the four trained classes.

Share X Bluesky LinkedIn Reddit HN

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 is a clear and honest assembly of known pieces: a federated autoencoder for anomaly detection, a LoRA-fine-tuned Llama 3.2 for threat classification, and mTLS/MQTT with automatic iptables enforcement. The testbed with five Raspberry Pis and 25 federated clients is real, and the resource measurements (inference latency, CPU/RAM, communication overhead) are useful. The related work is adequate, and the writing is readable. I believe the authors actually built this system.

The problem is the evaluation. The stress-test note is correct and lands squarely. Attack samples in Section IV.A.2 are generated by modifying features of a synthetic baseline built from the same normal traffic distribution; virtual devices are bootstrapped from the pooled dataset D_G (Eq. 3) with multiplicative shifts and noise. The threshold τ is the 95th percentile of the reconstruction MSE computed from all normal samples, and no train/test split is described. Under that construction, an autoencoder fitted on D_G will naturally assign higher reconstruction error to samples deliberately moved away from D_G. Perfect separation (AUC 1.0) and perfect recall are forced by the generation procedure, not measured on real attacks. The LLM evaluation inherits the same pipeline: the fine-tuning data and the test taxonomy are the same four attack classes, and there is no held-out set or real-world attack trace. Without baselines or external datasets, the central claims are unfalsifiable in the current form.

The numbers also have internal inconsistencies: the text says an F1-score of 0 in one place, the table says 0.909, the abstract says 0.9091, and Table IV gives an overall F1 of 0.9231. The 41-second LLM latency is worth flagging for real-time enforcement, though the enforcement step itself is fast.

These flaws are load-bearing. As a prototype description, the paper has value; as a research claim about a privacy-preserving zero-trust IoT defense, it fails. The authors should redo the evaluation with real attack traffic or a properly held-out public dataset, add baselines, and fix the inconsistency. I would not cite the numbers, and I would not send this to peer review in its current form, since the circularity is identifiable at the desk and the revision needed is substantial.

Referee Report

4 major / 6 minor

Summary. The paper proposes a zero-trust IoT architecture that combines federated learning (FL) with an autoencoder-based anomaly detector, a LoRA fine-tuned LLM (Llama 3.2-3B) for threat classification and response generation, and automated iptables enforcement on an MQTT/TLS broker. The system is evaluated on a testbed of five Raspberry Pis extended with statistically bootstrapped virtual devices. The paper reports an F1 score of 0.9091, ROC-AUC of 1.0, perfect recall across all attack classes, and perfect LLM classification accuracy, and claims that this demonstrates effective privacy-preserving anomaly detection for resource-constrained IoT devices.

Significance. If the reported results were valid, the architecture would be a useful integration of FL, LLM reasoning, and zero-trust enforcement for IoT. However, the empirical evaluation is constructed in a way that makes the headline performance essentially an artifact of the data-generation procedure. Attacks are synthetic modifications of the same normal-traffic distribution used to train the autoencoder and set the anomaly threshold; virtual devices are bootstrapped from the same pooled dataset; and no independent train/test split, real attack traces, or baselines are provided. The contribution is therefore currently a plausible system design rather than a validated detection framework. The paper does not provide code or data, so the results are not independently reproducible.

major comments (4)

[§IV.A.2, §IV.B] The central detection claim (ROC-AUC 1.0, recall 1.0, F1 0.9091) is an artifact of the evaluation design. Attack traffic is created by taking a synthetic baseline of normal MQTT/TLS traffic and modifying selected features to emulate malicious behavior. The autoencoder is trained on normal traffic, and the threshold τ is set as the 95th percentile of reconstruction MSE over normal samples. Under this construction, any deterministic feature perturbation moves attack samples away from the training manifold, guaranteeing high reconstruction error and near-perfect separation. No independent attack dataset or held-out normal set is used. The reported per-class F1 of 0.909 is simply a consequence of the fixed 5% false-positive rate and the balanced one-vs-rest evaluation, not a measured property of the method. The evaluation must be redone with real attack traces or a standard public IoT attack
[§IV.A.3, Eqs. (1)–(6)] Virtual devices are generated by bootstrapping the pooled dataset D_G and applying per-device multiplicative shifts and Gaussian noise. Because D_G is the same data used to train the autoencoder and to compute the threshold, the bootstrapped evaluation samples are near-duplicates of training samples. No device-wise, temporal, or other independent split is described. This creates a serious risk of data leakage: the reported FL convergence, resource measurements, and anomaly-detection scores may reflect memorization of training samples rather than generalization to genuinely new devices. The authors must describe an independent split and quantify the overlap between bootstrapped samples and training samples.
[§IV.D, Table IV] The LLM is fine-tuned on 500 instruction-response pairs generated for the same four attack classes used in the evaluation (TLS downgrade, DDoS SYN flood, port scan, plain MQTT) and then evaluated on anomalies produced by the same synthetic generator. Perfect precision and recall on a self-generated, in-distribution test set do not establish that the LLM can classify real attacks or handle out-of-distribution anomalies. At minimum, an evaluation on held-out attack descriptions, a different attack taxonomy, or a public dataset is required. The current Table IV is not informative about generalization.
[§IV.B] No baseline or ablation is reported. The paper does not compare the proposed FL autoencoder against a centralized autoencoder, an isolation forest, a supervised classifier, or a non-federated variant. Similarly, the LLM component is not compared with a generic prompted LLM or with a simpler rule-based classifier. Without such comparisons, the specific contributions of FL and the LLM are not quantified, and the resource-overhead tables (Tables II and III) cannot be interpreted. The paper should include baselines and ablations to support the claim that the proposed framework is effective.

minor comments (6)

[§IV.B, Table I] In the text, the F1-score is reported as '0' for each attack class ('F1-score of 0'); this is clearly a typo and should be 0.909.
[Abstract, Table I] The abstract reports F1 = 0.9091, which is the per-class macro average, while Table I also reports an Overall F1 of 0.976. Please clarify which F1 is meant in the abstract and conclusions.
[§III.B and §III.C] Feature engineering states that the feature space is reduced to 22 discriminative features, but the autoencoder is described with an input dimension of d=36. This inconsistency should be resolved; if one-hot expansions or additional features are used, they should be described explicitly.
[§IV.A] The introduction mentions 25 heterogeneous clients, while the physical testbed consists of five Raspberry Pis. Please clarify how the virtual devices map to the 25 clients and whether each virtual device corresponds to a separate FL client.
[§IV.B, Tables II–III] Resource measurements are reported as single averages without variance or confidence intervals. Since virtual-device generation is stochastic, reporting the spread across multiple random seeds would substantially improve interpretability.
[§IV.D] The end-to-end LLM latency of about 41 seconds relies on a Google Colab backend and an ngrok tunnel. This is a deployment limitation, not a property of the proposed architecture; it should be stated as such, and the feasibility of real-time enforcement should be discussed.

Circularity Check

2 steps flagged · score 8.0 of 10

Evaluation is circular: synthetic 'attacks' are perturbations of the same normal data used to train the autoencoder and set τ, so perfect AUC/recall and LLM accuracy are artifacts of the generation procedure.

fitted input called prediction [§IV.A.2-3 (Eqs. 1-6), §III.C, §IV.B]
"We first created a synthetic baseline representing normal MQTT/TLS IoT traffic and subsequently modified the selected network and protocol features to emulate malicious behavior. ... The anomaly detection threshold τ was defined as the 95th percentile of the reconstruction error distribution computed from all normal traffic samples. ... samples are first drawn using bootstrap sampling as in Eq. 3. x_boot_{j,f} ∼ D_G ... The final generated feature value is computed as in Eq. 5: ~x^{(v)}_{j,f} = clip( x_boot_{j,f} · (1 + δ^{(v)}_f) + η^{(v)}_{j,f}, 0, 1)."

The normal traffic data are collected from the Pi testbed and pooled into D_G; virtual devices are bootstrapped from D_G (Eq. 3) with only bounded multiplicative shifts (Eq. 2) and small Gaussian noise (Eq. 4). The attack samples are then made by modifying the same synthetic normal baseline. The anomaly threshold τ is defined as the 95th percentile of reconstruction error 'computed from all normal traffic samples' (§III.C, §IV.B). Thus the normal evaluation distribution is the same distribution used to fit τ, and 'attacks' are deliberately moved away from that distribution. An autoencoder trained on D_G will necessarily assign higher reconstruction error to points displaced from D_G, so ROC-AUC 1.0 and perfect recall are forced by the construction rather than measured on independent real a
other [§IV.A.4, §IV.B Table IV]
"The fine-tuning dataset consisted of 500 instruction–response pairs, balanced equally across the four attack classes: TLS downgrade, DDoS SYN flood, port scan, and plain MQTT (125 examples each). ... The LLM achieved perfect precision and recall (1.0000) across all four attack classes, correctly identifying the attack category and severity level for every anomalous sample."

The LLM is fine-tuned on exactly the same four attack classes on which it is evaluated, with 125 examples of each class, and the evaluation anomalies come from the same synthetic MQTT/TLS modification pipeline. No held-out attack class, no independently captured malicious traffic, and no distribution shift is reported. 'Perfect' classification on the training taxonomy is therefore an in-distribution measurement (essentially a memorization check), not a prediction of generalization to novel threats.

full rationale

The paper's headline results (F1 0.9091, ROC-AUC 1.0, perfect recall, and LLM perfect classification) are produced by a closed synthetic loop. Normal data are collected from Raspberry Pis and pooled as D_G; virtual devices are generated by bootstrapping D_G with small bounded noise (Eqs. 3-5); attack samples are created by modifying a synthetic normal baseline; and the anomaly threshold τ is the 95th percentile of reconstruction error over that same normal distribution. Consequently, any attack generated by moving features away from D_G will exceed τ by design, making perfect separation an artifact. The LLM reasoning layer is fine-tuned on 500 examples split equally across exactly the four attack classes it is later asked to classify, so Table IV's 1.0000 scores are expected from training-distribution evaluation. There is no external dataset, no real attack trace, and no held-out distribution for either the autoencoder or the LLM. The circularity is in the evaluation construction, not in the FL/LLM methodology itself; a real-world or independent benchmark could break the loop. No self-citation chain is involved, but the central performance claims reduce to the data-generation and threshold-fitting procedure by construction.

Assumptions & free parameters 4 free parameters · 4 assumptions · 1 invented entities

The central evaluation rests on three unvalidated choices: synthetic attack construction, bootstrapped virtual devices, and a p95 threshold fitted to the same normal data. Any one of these being unrealistic would invalidate the reported 1.0 AUC and perfect recall.

free parameters (4)

Anomaly threshold τ (95th percentile of reconstruction MSE) = 95th percentile over normal training data
Section III.C: this fitted threshold sets the 5% false-positive rate and directly shapes the reported precision/F1 values.
Bootstrap noise scale σ_noise,f = max(0.02·σ_f,G, 1e-6)
Eq. 4: hand-chosen scale controls virtual-device feature variability and influences any evaluation performed on those devices.
CV-based multiplicative shift range for virtual devices = U(-CV_f,G, CV_f,G) from real device stats
Eq. 2: defines allowable deviation of 20 virtual clients; no independent validation that this mimics real hardware diversity.
Feature reduction to 22 features
Section III.B: the particular six continuous and sixteen binary features are chosen by hand; the exact list is not provided, and all downstream results depend on it.

assumptions (4)

ad hoc to paper Attack samples generated by modifying features of normal MQTT/TLS traffic are representative of real attacks.
Section IV.A.2: synthetic baseline modified to emulate port scan, DDoS SYN flood, TLS downgrade, and plain MQTT; the autoencoder threshold is set on the same normal distribution.
ad hoc to paper CV-based bootstrap virtual devices (Eqs. 1-6) are statistically representative of real heterogeneous IoT devices.
Section IV.A.3: 20 of 25 clients are generated by resampling real traffic with hand-set noise and shift ranges; no external validation.
domain assumption The 95th percentile reconstruction-error threshold yields valid anomaly detection.
Section III.C: assumes the normal reconstruction-error distribution is stable and that all real attacks produce higher MSE than this percentile.
domain assumption The LoRA fine-tuned Llama 3.2-3B can map feature-vector prompts to correct attack classes and severities.
Sections III.D and IV.A.4: evaluation uses the same four attack classes as the 500-sample fine-tuning set, with no distinct held-out test described.

invented entities (1)

Virtual IoT devices generated by CV-based bootstrap
purpose: Expand the testbed from five physical Raspberry Pis to 25 heterogeneous FL clients and produce the majority of the 1.9M evaluation samples.
Eqs. 1-6 create synthetic devices by resampling and perturbing real traffic; there is no external check that these devices behave like real hardware, and the resulting test data contaminates the evaluation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Federated Learning and LLM-Driven Threat Intelligence for Zero Trust IoT Architecture." pith.science (2026). https://pith.science/paper/H22DNXET

@misc{pith2026260717035,
  author       = {Pith},
  title        = {Pith review of: Federated Learning and LLM-Driven Threat Intelligence for Zero Trust IoT Architecture},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H22DNXET}},
  note         = {Machine review of arXiv:2607.17035}
}

read the original abstract

While the Internet of Things (IoT) has become essential, they introduced serious security and privacy challenges, especially for mission-critical environments. Legacy devices are vulnerable to viruses, data breaches, and unauthorized access, and updating these devices would be infeasibly costly. As a solution, this paper presents a Federated Learning and LLM-Driven Threat Intelligence for Zero Trust IoT Architecture, with FL for anomaly detection integrating privacy-preserving distributed learning, continuous identity verification, and LLM-driven autonomous threat response into a unified pipeline. Unlike existing solutions, our framework enforces Zero Trust at every communication layer via mutual TLS (mTLS) over MQTT, ensuring no device or message is implicitly trusted. Our experiments with Raspberry Pis and various sensors achieve an F1 score of 0.9091 and an ROC-AUC of 1.0, highlighting the effectiveness of the proposed framework in enabling privacy-preserving anomaly detection for resource-constrained IoT devices.

Figures

Figures reproduced from arXiv: 2607.17035 by the authors.

**Figure 2.** FL-based Autoencoder architecture. After training, the global model is evaluated on normal datasets to compute the anomaly detection threshold τ . We set τ as the 95th percentile of the reconstruction MSE distribution over normal data, providing a 5% theoretical false positive rate. Any packet exceeding τ is classified as ANOMALY and forwarded to the LLM for analysis. D. LLM-Based Threat Intelligence Generation The … view at source ↗

**Figure 3.** FL convergence of the global autoencoder across 10 comm. rounds. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: End-to-end ZT-FL enforcement pipeline. V. CONCLUSION This work demonstrates that combining FL with LLMdriven reasoning enables a practical, scalable, and fully autonomous Zero Trust defense for IoT environments, capable of detecting, interpreting, and mitigating thr…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 1 linked inside Pith

[1]

S. Sinha. (2025) State of iot 2025: Number of connected iot devices growing 14% to 21.1 billion globally. IoT Analytics. Accessed: April 6, 2026. [Online]. Available: https://iot-analytics.com/ number-connected-iot-devices/

2025
[2]

Internet of Things (IoT) market size, share & industry analysis, by component, by deployment, by enterprise type, by industry, and regional forecast, 2026–2034,

Fortune Business Insights, “Internet of Things (IoT) market size, share & industry analysis, by component, by deployment, by enterprise type, by industry, and regional forecast, 2026–2034,” Fortune Business Insights, Tech. Rep. FBI100307, May 2026, accessed: May

2026
[3]

(2026) Iot hacking statistics 2026 — global trends and key insights

Knowledge Hub. (2026) Iot hacking statistics 2026 — global trends and key insights. DeXpose. Accessed: April 6, 2026. [Online]. Available: https://www.dexpose.io/iot-hacking-statistics/

2026
[4]

Zero trust ar- chitecture,

S. Rose, O. Borchert, S. Mitchell, and S. Connelly, “Zero trust ar- chitecture,” National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA, Tech. Rep. NIST Special Publication 800- 207, 2020

2020
[5]

Communication-efficient learning of deep networks from decentralized data,

B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” inArtificial intelligence and statistics. Pmlr, 2017, pp. 1273– 1282

2017
[6]

Future industry internet of things with zero-trust security,

S. Li, M. Iqbal, and N. Saxena, “Future industry internet of things with zero-trust security,”Information Systems Frontiers, vol. 26, no. 5, pp. 1653–1666, 2024

2024
[7]

Zero-trust hierarchical management in iot,

M. Samaniego and R. Deters, “Zero-trust hierarchical management in iot,” in2018 IEEE International Congress on Internet of Things (ICIOT), 2018, pp. 88–95

2018
[8]

Flexible zero trust architecture for the cybersecurity of industrial iot infrastructures,

C. Zanasi, S. Russo, and M. Colajanni, “Flexible zero trust architecture for the cybersecurity of industrial iot infrastructures,”Ad Hoc Networks, vol. 156, p. 103414, 2024

2024

Show all 15 references

[9]

Recent advances on federated learning for cybersecurity and cybersecurity for federated learning for internet of things,

B. Ghimire and D. B. Rawat, “Recent advances on federated learning for cybersecurity and cybersecurity for federated learning for internet of things,”IEEE Internet of Things Journal, vol. 9, no. 11, pp. 8229–8249, 2022

2022
[10]

Federated-learning-based anomaly detection for iot security attacks,

V . Mothukuri, P. Khare, R. M. Parizi, S. Pouriyeh, A. Dehghantanha, and G. Srivastava, “Federated-learning-based anomaly detection for iot security attacks,”IEEE Internet of Things Journal, vol. 9, no. 4, pp. 2545–2554, 2021

2021
[11]

Federated learning for malware detection in iot devices,

V . Rey, P. M. S. S ´anchez, A. H. Celdr ´an, and G. Bovet, “Federated learning for malware detection in iot devices,”Computer networks, vol. 204, p. 108693, 2022

2022
[12]

An llm-based framework for finger- printing internet-connected devices,

A. Sarabi, T. Yin, and M. Liu, “An llm-based framework for finger- printing internet-connected devices,” inProceedings of the 2023 ACM on Internet Measurement Conference, 2023, pp. 478–484

2023
[13]

Llm-based threat detec- tion and prevention framework for iot ecosystems,

Y . Otoum, A. Asad, and A. Nayak, “Llm-based threat detec- tion and prevention framework for iot ecosystems,”arXiv preprint arXiv:2505.00240, 2025

2025 arXiv
[14]

Security in the internet of things: A survey on application layer protocols,

L. Nastase, “Security in the internet of things: A survey on application layer protocols,” in2017 21st International Conference on Control Systems and Computer Science (CSCS), 2017, pp. 659–666

2017
[2026]

Available: https://www.fortunebusinessinsights.com/ industry-reports/internet-of-things-iot-market-100307

[Online]. Available: https://www.fortunebusinessinsights.com/ industry-reports/internet-of-things-iot-market-100307

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

S. Sinha. (2025) State of iot 2025: Number of connected iot devices growing 14% to 21.1 billion globally. IoT Analytics. Accessed: April 6, 2026. [Online]. Available: https://iot-analytics.com/ number-connected-iot-devices/

2025

[2] [2]

Internet of Things (IoT) market size, share & industry analysis, by component, by deployment, by enterprise type, by industry, and regional forecast, 2026–2034,

Fortune Business Insights, “Internet of Things (IoT) market size, share & industry analysis, by component, by deployment, by enterprise type, by industry, and regional forecast, 2026–2034,” Fortune Business Insights, Tech. Rep. FBI100307, May 2026, accessed: May

2026

[3] [3]

(2026) Iot hacking statistics 2026 — global trends and key insights

Knowledge Hub. (2026) Iot hacking statistics 2026 — global trends and key insights. DeXpose. Accessed: April 6, 2026. [Online]. Available: https://www.dexpose.io/iot-hacking-statistics/

2026

[4] [4]

Zero trust ar- chitecture,

S. Rose, O. Borchert, S. Mitchell, and S. Connelly, “Zero trust ar- chitecture,” National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA, Tech. Rep. NIST Special Publication 800- 207, 2020

2020

[5] [5]

Communication-efficient learning of deep networks from decentralized data,

B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” inArtificial intelligence and statistics. Pmlr, 2017, pp. 1273– 1282

2017

[6] [6]

Future industry internet of things with zero-trust security,

S. Li, M. Iqbal, and N. Saxena, “Future industry internet of things with zero-trust security,”Information Systems Frontiers, vol. 26, no. 5, pp. 1653–1666, 2024

2024

[7] [7]

Zero-trust hierarchical management in iot,

M. Samaniego and R. Deters, “Zero-trust hierarchical management in iot,” in2018 IEEE International Congress on Internet of Things (ICIOT), 2018, pp. 88–95

2018

[8] [8]

Flexible zero trust architecture for the cybersecurity of industrial iot infrastructures,

C. Zanasi, S. Russo, and M. Colajanni, “Flexible zero trust architecture for the cybersecurity of industrial iot infrastructures,”Ad Hoc Networks, vol. 156, p. 103414, 2024

2024