Model verification for population detection of counterfeits.

Area of Science:

• Forensic Science
• Manufacturing Process Analysis
• Quality Control

Background:

• Counterfeit item quality has significantly increased due to advanced global manufacturing capabilities.
• Authenticating manufactured goods, including coinage, relies on analyzing inherent product features.
• Detecting fraud involves comparing an 'Example' group against a 'Questioned' group using established testing protocols.

Purpose of the Study:

• To validate and enhance a previously reported counterfeit detection model using reconstruction techniques.
• To improve model accuracy and confirm previous conclusions regarding fraudulent items.
• To demonstrate the effectiveness of analyzing manufacturing process variation for fraud detection.

Main Methods:

• Developing predictive models for both 'Example' and 'Questioned' groups based on standard tests of individual items, using coin weight as an illustrative example.
• Applying reverse-quality-engineering methods to analyze variations between individual pieces and determine process capability.
• Utilizing reconstruction techniques to re-create the evidence set for model validation.

Main Results:

• The analysis confirmed that the 'Questioned' set of coins was likely over 37% non-conforming by weight.
• Manufacturing process capability, rather than adherence to specifications alone, can differentiate authentic from fraudulent items.
• The study demonstrated that integrating additional analytical methods beyond the original modeling software improved decision outcomes.

Conclusions:

• The proposed method effectively identifies fraudulent items by analyzing manufacturing process variations.
• Reverse-quality-engineering and analysis of process capability are powerful tools for counterfeit detection.
• Enhanced analytical approaches can significantly improve the accuracy and reliability of fraud detection models.