Technical Stream
Technical Stream
Brand logo recognition system

Abstract:

Conduct research in the field of pattern/text recognition and develop software capable of identifying the name of a brand/company/application from an image containing the logo of a famous brand, company or application.

Objective:

Cybercriminals often use a social engineering technique known as clone phishing. This involves the partial or total copying of a legitimate object’s attributes and using these attributes to both conceal malicious activity and trick the user. This technology is used to create phishing pages for social networking sites and banks, phishing emails including the logos of payment systems, or malicious mobile applications with a similar interface to the original. The ultimate goal of phishing, as a rule, is to steal user logins/passwords.

To create an effective automated system for analyzing cloned objects, it is necessary to identify the initial object which has been cloned. One method of solving the problem is to create a neural network that recognizes the images of well-known brands and that can be trained using a specially created set of samples. The system works according to the scheme shown in Figure 1 below.

1. Brand Logo Recognition System

Figure 1

The aim of this study task is to analyze image-recognition methods and create a system capable of identifying the brands assigned to it from an image that is sent as input, and provide their names in text form.

Data:

The researchers are provided with a set of screenshots of phishing websites, email and mobile applications in JPEG.

The data is available after registration.

Requirements:

To analyze possible methods for solving the task and to implement an effective software prototype.

Evaluation criteria:

1) The percentage of correctly recognized brands

2) Scalability of the application

3) Performance: more than 1 image per second