Vulnerability analysis of neural networks in computer vision

The work considers the actual problem of the vulnerability of artificial intelligence technologies based on neural networks. We show that the use of neural networks generates many vulnerabilities. We demonstrate specific examples of such vulnerabilities: incorrect classification of images containing adversarial noise or patches, failure of recognition systems in the presence of special patterns on the image, including those applied to objects in the real world, training data poisoning, etc. Based on the analysis, we show the need to improve the security of artificial intelligence technologies and suggest some considerations that contribute to this improvement.