Fast detection of fake and tampered cards with artificial intelligence

José Luis Domínguez
3 min readApr 9, 2022

AI + OpenCV

Background

Photo by Pickawood on Unsplash

Tax fraud is a persistent fact of everyday and business life, affecting companies of all sizes, various industries and society. Although there is no infallible method to prevent fraud, the risk can be minimized by structuring an early prevention system in the face of these events.

Description of the problem

The manipulation of bank cards and personal identification has become a major concern for companies and societies around the world. Every year, countless people and organizations are victims of fraud and monetary loss. Since cyber attackers use different tactics to steal or spoof information.

How does Artificial Intelligence participate as a solution?

Photo by Markus Spiske on Unsplash

Given this situation, the need to configure advanced identity verification solutions to safeguard personal information is identified. The detection of manipulation and fraud through computer vision and artificial intelligence represents the most timely and effective solution to these attacks of information theft.

What problem should AI solve?

How to demonstrate that the use of computer vision, a subfield of artificial intelligence, can minimize the risk of manipulation of bank cards and personal identification.

Data Scientist Goals

  1. Detect the manipulation of information in bank cards and personal identification through computer vision.

2. Evaluate the measurement and prediction of non-structural information, through an initial image without compression or without distortions as a reference.

Answer the question: Are computer vision methods the most suitable for monitoring and evaluating early detection of counterfeit and tampered cards?

Type of research and development

Quantitative research, since the vectorization of pixels to numerical values is considered, to later be processed in computer vision algorithms (Structural Similarity Index) that evaluate the texture, luminosity, structure, filters and detection of characteristics.

The Structural Similarity Index (SSIM) is based on centralization types of neighborhood pixels (Figure 1) in various windows of an image (Figure 2). Which is the product between two partial indices (x, y) of common size N × N. The first index SL contains local luminance and the second SV covariance / local variance (Figure 3).

This means that the statistics of the mirror image (image to be contrasted) are simply related to the original image, which can be directly interpreted as a luminance (square) contrast and a variance contrast.

Figure 1. Centralization of neighborhood pixels in an image.
Figure 2. Directional windows in a map montage for a given image.
Figure 3. Formula of the Structural Similarity Index (SSIM).

Workflow Development

Results

--

--

José Luis Domínguez

Data scientist who develops sustainable reliability in the processes driven by the development of artificial intelligence in future society.