Information Security of Neural Networks in the Russians’ Assessments: From Professional to Personal Practices

The article analyzes the Russians’ assessments of information security issues related to neural networks amid a shift in the emphasis of their use from professional practices to personal ones, while parallel or hybrid use is possessed. Drawing on OLS regression, the relationship between the intensity of neural network usage and the Russians’ assessments of neural network security is identified. With large language models as a case study, the Kruskal–Wallis test has been chosen to appoint a significance of the differences in assessments of their information security across the groups distributed by gender, age, and education level. The obtained results testify to the uniformity of critical attitude towards the information security apart from gender or age and the presence of statistically significant differences by education level. A contradiction between the adoption of large language models as a tool and the perception of low user security is highlighted. The study notes the advisability of embedding user protection mechanisms directly into the neural network interfaces.