Unveiling scientific articles from paper mills with provenance analysis

Area of Science:

• Scientific integrity
• Bibliometrics
• Image analysis

Background:

• Paper mills generate fake publications, challenging scientific integrity.
• Current methods for identifying fraudulent articles are often manual and inefficient.
• Fake publications frequently use duplicated or manipulated figures for dissemination.

Purpose of the Study:

• To develop an automated methodology for identifying systematically produced fraudulent publications.
• To group manuscripts originating from the same paper mill based on figure analysis.
• To provide a tool for enhancing scientific integrity by detecting paper mill activities.

Main Methods:

• Developed a provenance analysis methodology to track and group suspicious figures and documents.
• Implemented figure analysis to identify duplicated and manipulated regions across manuscripts.
• Constructed a provenance graph to link and organize evidence of systematic production.

Main Results:

• Successfully grouped systematically produced articles from paper mills on datasets of varying sizes.
• Identified reused and manipulated figures as key indicators of fraudulent manuscript production.
• Demonstrated the method's effectiveness even when faced with intentionally distracting data.

Conclusions:

• The proposed provenance analysis technique offers a promising automated solution for detecting fraudulent manuscripts.
• This method can effectively identify articles produced by paper mills by analyzing figure reuse and manipulation.
• The tool supports scientific integrity by providing evidence of systematic fraudulent publication practices.