Intelligent System for Preventing Rubber Ducky Attacks Using Deep Learning Neural Networks

Abstract

At the present stage of development of hardware and software in the field of information technology an important place is occupied by information security, which implies the use of additional means to ensure secure processing of confidential data of end-users of target systems. The main problem, allocated in existing electronic computers, is insufficient protection of hardware ports from external influences. An example of such vulnerabilities is the USB Rubber Ducky hardware and software security solution from hak5 that exploits vulnerabilities by emulating peripherals to perform unauthorised actions on the target machine. Offensive Security's advanced, similar solutions allow complex target machine interception and remote control of the target machine, underscoring the urgency of the problem at hand. To solve this problem, an intelligent system was developed to evaluate and calculate the states of connected peripheral devices, in particular, to evaluate the comparability of input groups of peripheral device parameters. To improve security, a deep learning artificial neural network process was integrated into the implemented system. Based on user actions – software processing call speed, user errors and existing exploitation patterns of variant vulnerabilities – it can identify a device as potentially dangerous and then hardware disconnect the USB port. The artificial neural network's learning functionality based on user behaviour patterns also allowed for personal identification without an active account, which has a positive impact on system security. An important feature of the system is also the ability to interact with the system remotely using the Telegram API.