Модели и алгоритмы защиты информации с использованием генеративно-состязательных нейронных сетей и данных о пользователе
Аннотация
В работе автор напоминает об актуальности проблемы информационной безопасности в условиях цифровизации и активного развития технологий машинного обучения. Также, автор подчеркивает популярность использования пользователями слабых паролей, содержащих личную информацию о пользователе, которые уязвимы для взлома. В работе рассмотрены существующие методы и подходы для генерации паролей и проведения их аудита. Для эффективного выявления небезопасных паролей автор описывает и предлагает вариант подхода к генерации списков паролей на основе генеративно-состязательной нейронной сети и известных данных о пользователе (таких как ФИО, дата рождения, увлечения, ФИО родственников и прочее), который может быть использован при проведении аудита информационной безопасности для выявления уязвимых паролей, проведения тестирования на проникновение, а также для решения некоторых задач при проведении судебно-компьютерной экспертизы. Описана и представлена модель GAN-сети. Для обучения нейронной сети предлагается использовать общедоступные данные из открытых источников, которые стали известны в результате различных утечек персональных данных, включающие в себя скомпрометированные пароли, а также информацию из различных поисковых систем и социальных сетей, таких как «Вконтакте», «Одноклассники» и др.
In this article, the author recalls the relevance of the problem of information security in the context of digitalisation and active development of machine learning technologies. information security in the context of digitalisation and active development of machine learning technologies. machine learning technologies. Also, the author emphasises the popularity of the use of weak passwords containing personal information about the user. weak passwords containing personal information about the user, which are vulnerable to hacking. to hacking. The paper considers existing methods and approaches for generating passwords and their auditing. To effectively identify insecure passwords, the author describes and proposes a variant of the approach to generating password lists based on the generative-adversarial neural network and known data about the user (such as full name, date of birth, hobbies, relatives' names, etc.), which can be used in information auditing. can be used in information security audits to identify vulnerable passwords, perform penetration testing, and also for solving some tasks in forensic computer forensics. A GAN model is described and presented the model of GAN-network. To train the neural network it is proposed to to use publicly available data from open sources, which became known as a result of various leaks of personal data, including compromised passwords, as well as information from various search engines and social networks, such as Vkontakte. social networks such as Vkontakte, Odnoklassniki and others.
In this article, the author recalls the relevance of the problem of information security in the context of digitalisation and active development of machine learning technologies. information security in the context of digitalisation and active development of machine learning technologies. machine learning technologies. Also, the author emphasises the popularity of the use of weak passwords containing personal information about the user. weak passwords containing personal information about the user, which are vulnerable to hacking. to hacking. The paper considers existing methods and approaches for generating passwords and their auditing. To effectively identify insecure passwords, the author describes and proposes a variant of the approach to generating password lists based on the generative-adversarial neural network and known data about the user (such as full name, date of birth, hobbies, relatives' names, etc.), which can be used in information auditing. can be used in information security audits to identify vulnerable passwords, perform penetration testing, and also for solving some tasks in forensic computer forensics. A GAN model is described and presented the model of GAN-network. To train the neural network it is proposed to to use publicly available data from open sources, which became known as a result of various leaks of personal data, including compromised passwords, as well as information from various search engines and social networks, such as Vkontakte. social networks such as Vkontakte, Odnoklassniki and others.