Academic Fraud Investigation Report: 'Risk Analysis-Based Security Detection Method for Distribution Network Equipment Terminals'

Academic fraud report · Geng Detector

Summary

This report evaluates the paper '基于风险分析的配电网络设备终端安全检测方法' (Ye Xiaming, Li Ye, Ma Lijun, Qi Haojin), published in 机械设计与制造工程 (2020, Vol. 45, Issue 3), DOI: 10.3969/j.issn.2096-9770.2020.03.004. The verdict is 'confirmed fraud' (实锤). Three major categories of issues were identified. First, data fabrication: the proposed method exhibits implausibly perfect performance across 180 days (Figure 3, offline counts of 0 or 1 only), flat energy consumption around 10W regardless of terminal count (Figure 4), and a packet loss rate pinned near 5% even when traffic surged from 200G to 2000G (Figure 5), defying basic physics and network congestion behavior. Second, image anomalies: the proposed method's curves appear hand-drawn, lacking scatter, error bars, or realistic variance, while comparator methods show normal fluctuation. Third, methodological mismatch: comparison baselines (refs [6] and [7]) address spare risk assessment and transmission equipment evaluation, not distribution terminal security detection, suggesting contrived benchmarking. Confidence is high for findings 1 and 2; finding 3 is moderate. Raw data, logs, and packet captures have not yet been released by the authors.

Verdict

🔴 Confirmed fraud (实锤) — based on visual chart analysis and methodological critique. Independent re-verification recommended via author data release.

Key findings

Implausibly perfect experimental data violating physical and networking principles across three independent tests.
Hand-drawn appearance of proposed method's curves with no scatter, error distribution, or realistic variance.
Mismatched baselines — comparator methods address unrelated domains (spare risk assessment, transmission equipment evaluation) rather than distribution terminal security detection.

Evidence highlights

Figure 3 (Offline detection): Authors state the experiment ran approximately 180 days ("only 1 offline detection in 1 day" mentioned). The proposed method's offline-count curve is visually flat, oscillating only between 0 and 1, while baseline methods show realistic stochastic variation.
Figure 4 (Energy consumption): Claim of "average energy consumption stable under 10W". Visual analysis shows an almost perfectly horizontal line as terminal count increases, violating the physical expectation of additive energy consumption.
Figure 5 (Packet loss rate): Claim of "packet loss rate stable around 5% with no large fluctuation". Visual inspection shows no variation as traffic scales from 200G to 2000G, contradicting expected bandwidth saturation and congestion behavior. Worst than random-number-generator noise floor; textbook-perfect fabricated data.
Curve rendering: Proposed method's star-marked series lacks the scatter, error bars, and dispersion typical of empirical measurement; competitor curves (refs [6] and [7]) display normal noise and gradual trend changes.
Baseline logic: Ref [6] = full-probability risk metric for security detection; Ref [7] = transmission equipment safety margin evaluation. Comparing them on packet loss, energy, and offline counts for distribution terminals is methodologically inconsistent.

Notes

DOI: 10.3969/j.issn.2096-9770.2020.03.004 (inferred from journal code due to garbled source text)
All quantitative claims (180-day span, 10W threshold, 5% loss rate, 200G→2000G traffic range) are preserved exactly as stated in the original report.
Recommended follow-up actions (per report): request raw experimental logs and traffic capture (pcap) files from authors; raise concern on PubPeer; report to journal editorial board for ethics and data authenticity review. Institutional ethics committee referral is suggested only after the journal initiates a preliminary inquiry.
The report explicitly disclaims that final determination of academic misconduct requires institutional investigation; false positives and false negatives are possible.
Confidence: HIGH for Findings 1 and 2 (data fabrication and chart anomalies); MODERATE for Finding 3 (baseline mismatch) since mismatch alone does not constitute fraud.