Integrity Review Report: "Research on IoT Anomaly Detection and SDN Threat Mitigation Based on Edge Computing" (DOI: 10.19304/J.ISSN1000-7180.2025.0006)

Academic fraud report · Geng Detector

Summary

Verdict: Strong indicators of serious academic misconduct, with multiple fatal flaws in algorithmic logic, arithmetic, and internal consistency. The paper proposes an IoT anomaly-detection system that defines a Pearson correlation threshold of exactly 1.0 (declaring any value less than 1 as abnormal), while simultaneously claiming a false-positive rate of only 0.9%–1.2%—an irreconcilable contradiction. In Table 4, the mean response time of seven samples (5, 5, 8, 7, 5, 6, 5) is reported as 5 s rather than the arithmetic ~5.857 s, and the mean false-positive rate is claimed as 1.1% rather than ~0.943%. Headline accuracy is stated as 98.53% in the abstract/conclusion but 98.51% in the body text. The claim of "~3% improvement" over baselines (96.70% and 95.70%) is mathematically exaggerated. Confidence is high for the arithmetic and consistency errors; image-based checks were not possible because the source PDF contained no analyzable figure data.

Verdict

🔴 Confirmed serious issues — The paper displays fundamental logical contradictions, elementary arithmetic mistakes, and inconsistent headlining statistics. These collectively indicate fabricated or carelessly assembled data and warrant formal institutional investigation.

Key findings

Fatal logical error in anomaly detection logic (Section 3, p. 66): The Pearson correlation threshold is set to exactly 1, meaning only perfectly linearly proportional sensor pairs are labeled "normal." This makes the false-positive rate claim (0.9%–1.2%) impossible.
Elementary mean-computation errors in Table 4: Response-time mean claimed as 5 s instead of the correct ~5.857 s; false-positive-rate mean claimed as 1.1% instead of the correct ~0.943%. The reported "averages" match the first row of Table 4, indicating copy-paste substitution.
Inconsistent headline accuracy: Abstract and conclusion state 98.53%; body text and Table 3 comparison report 98.51%. A 0.02 percentage-point discrepancy in the paper's key selling metric.
Mathematically exaggerated comparison claim: Body text claims "approximately 3% improvement" over the two cited baselines, but the actual deltas vs. the reported references are ~1.81% and ~2.81% respectively.
Image-based forensic analysis was not possible because the source PDF contained no analyzable pixel-level images.

Evidence highlights

Table 4 response-time samples: 5, 5, 8, 7, 5, 6, 5 → sum 41, n=7 → 41/7 ≈ 5.857 s (paper states 5 s).
Table 4 false-positive-rate samples: 1.1, 1.2, 0.9, 0.3, 1.0, 0.9, 1.2 → sum 6.6, n=7 → 6.6/7 ≈ 0.943% (paper states 1.1%).
Quoted threshold definition: "Pearson correlation ρ' ∈ [−1, 1] … therefore set threshold = 1; only perfectly positively correlated data are judged normal; less than 1 is judged abnormal" — directly contradicting the claimed false-positive rate of 0.9%–1.2%.
Headline accuracy contradiction: 98.53% (abstract/conclusion) vs 98.51% (Table 3 / body).
Improvement claim: paper states "~3% improvement" over Literature [1] and [2]; reported baseline values 96.70% and 95.70% yield deltas of 1.81% and 2.81% respectively.
DOI: 10.19304/J.ISSN1000-7180.2025.0006.

Notes

The reviewer's arithmetic and logical checks are deterministic and reproducible from the reported data; no inference from unverifiable material was required.
Because the source PDF did not provide figure data of sufficient resolution, no image-manipulation forensic analysis was performed.
The findings herein are based solely on the internal text and tabulated data; final adjudication must come from the journal's editorial office and the authors' institution (China Tower Co., Ltd., Jiangsu Branch).