Handling consumer vulnerability in e-commerce product images using machine learning.

Area of Science:

• E-commerce security
• Machine learning applications
• Consumer protection

Background:

• Growing interest in secondhand products highlights consumer vulnerability to deceptive online practices.
• Existing research inadequately addresses issues related to product display images and seller credibility in e-commerce.

Purpose of the Study:

• To develop a novel method for detecting vulnerabilities in reused product images on e-commerce platforms.
• To enhance consumer confidence and reduce risks associated with purchasing secondhand goods online.

Main Methods:

• A three-step convolutional neural network approach is employed for image analysis.
• The method, termed product image-based vulnerability detection (PIVD), integrates image processing and machine learning techniques.
• PIVD is designed to identify fraudulent dealers by analyzing product images.

Main Results:

• The proposed PIVD method demonstrates superior performance compared to standard CNN and CNN-LSTM models.
• Achieved an F1 score 2.3% higher than CNN (various filter sizes), 4% higher than CNN-LSTM (learning rate 0.008), and 6% higher than CNN (dropout 0.5).

Conclusions:

• The developed PIVD method effectively addresses consumer concerns regarding secondhand online purchases.
• This approach empowers buyers with greater assurance and minimizes potential damages from fraudulent listings.