Fraud Investigation Report: A Fine-Grained IoT Data Access Control Scheme Combining Attribute-Based Encryption and Blockchain (DOI: 10.1155/2021/5308206)

Academic fraud report · Geng Detector

Summary

This report assesses the 2021 Hindawi paper by Xiaofeng Lu, Songbing Fu, Cheng Jiang, and Pietro Lio, published in Security and Communication Networks, for suspected data fabrication and scholarly carelessness. The overall verdict is highly suspicious. Three principal issues are identified. First, Figures 7 and 8 (upload/download times for the 'Ours' scheme) show a perfectly horizontal line across file sizes from 500KB to 1200KB, violating the physical expectation that larger files require more transmission time. The authors' own claim of fixed 8% and 11% ratios versus Wang's scheme is mathematically incompatible with one curve being constant and the other increasing. Second, Figure 6 (encryption/decryption times) shows an implausibly smooth curve despite being the average of 50 runs, lacking the noise typical of real cryptographic benchmarks. Third, the manuscript contains multiple careless citation errors, including misattributing ABE to Waters alone rather than Sahai and Waters, and misnaming the authors of reference [24]. Confidence is moderate; final determination requires original experimental logs and institutional investigation.

Verdict

🟠 Highly suspicious. Multiple indicators point to fabricated or non-empirically derived performance data, compounded by careless citation handling.

Key findings

Horizontal-line paradox in Figures 7 and 8: The "Ours" upload and download time curves appear as flat horizontal lines across file sizes 500KB–1200KB, contradicting basic network/storage physics where larger files must take longer.
Mathematical inconsistency with stated ratios: The text claims IPFS ("Ours") times are ~8% (upload) and ~11% (download) of the cloud-storage baseline. A constant numerator paired with a growing denominator cannot produce constant percentage ratios, indicating the figures and prose are decoupled.
Implausibly smooth Figure 6: Although Section 5.3.1 states results are averages of 50 runs, the encryption/decryption curve is overly smooth, with no visible noise or error bars typical of real CPU/memory-bound benchmarks.
Citation and authorship errors: Introduction attributes ABE to "Waters" alone in 2005, though reference [11] lists "A. Sahai and B. R. Waters"; Section 5.3.1 cites "Hao Wang and Song [24]" while reference [24] lists "H. Wang and Y. Song".

Evidence highlights

Figure 7 (Upload times) and Figure 8 (Download times): "Ours" curve is horizontal from 500KB to 1200KB; Wang [25] curve grows monotonically.
Stated ratios in Section 5.3.2: "about 8%" and "about 11%" — arithmetically incompatible with the plotted flat line.
Section 5.3.1 declaration: "average results of 50 experimental runs" — not reflected in the smoothness of Figure 6.
Reference [11]: "A. Sahai and B. R. Waters" vs. text's single attribution to Waters.
Reference [24]: "H. Wang and Y. Song" vs. text's "Hao Wang and Song".

Notes

DOI: 10.1155/2021/5308206
Severity ratings in original report: Finding 1 🔴, Finding 2 🟠, Finding 3 🟡.
Recommended actions in original report: request raw experimental logs for Figures 7 and 8, post concerns on PubPeer, and report to the journal editorial office.
The report explicitly disclaims that AI-assisted findings are not a definitive ruling; final determination requires institutional investigation and original data verification.