Module for searching destructive information in images

Abstract

Module for searching destructive information in images

Dzhurov A.A., Cherkesova L.V., Revyakina E.A.

Incoming article date: 27.06.2024

Images on web-sites, social networks, computers may contain destructive content and pose a threat to the psyche of a child or adolescent. Conventional image classification does not always classify it correctly and accordingly has a number of drawbacks due to which there can be false positives, which reduces the accuracy of classification. The paper presents a method in Python module that can detect malicious content in images. The method described in the paper is based on the use of Yolov8 library, which provides good classification of images and further analysis. Using the developed method it was possible to reduce the number of false positives, which led to an increase in its efficiency. The paper shows the scheme of operation of the new method, as well as demonstrated the search for objects in images. Similar programs are considered and their comparative analysis with the developed method is carried out.

Keywords: Spacy, disruptive images, information security, pymorphy3, ultralytics, disruptive text, disruptive content, YoloV8, child safety, benchmarking, digital hash