Scientists at the St. Petersburg Federal Research Center of the Russian Academy of Sciences have developed and trained a neural network to detect malicious programs known as keyloggers, designed to steal passwords. The use of this technology will significantly improve the network security system and protect users from unauthorized access.
Researchers at St. Petersburg Federal Research Center of the Russian Academy of Sciences have proposed an approach that uses neural networks to detect keyloggers — programs that record the sequence of keystrokes on a keyboard or mouse. The proposed algorithms can be integrated into network security systems to protect against attackers who may use keyloggers, for example, to gain access to user accounts.
Keyloggers are specialized programs or devices that record keystrokes on a keyboard and save them in a special file. This technology can be used for legitimate purposes, such as monitoring employee performance, as well as for illegal purposes.
The main danger posed by keyboard spies is that they can discreetly collect confidential data, which can lead to fraud, identity theft, and even financial losses. In addition, keyloggers can be part of more complex malicious programs that use the collected data for further attacks, such as phishing or virus distribution.
We have developed an approach that looks for traces of keyloggers in network traffic, that is, it targets the process of interaction between spyware and remote servers. Our solution is based on several artificial intelligence methods that can monitor user or organization traffic and signal if suspicious network activity similar to keylogger activity is detected.
The research center noted that during the experiments, scientists studied publicly available datasets that included information about the operation of keyboard spies. After preliminary data processing, they selected and tested various machine learning models. The models were evaluated according to several criteria: how well they detect keyboard spies and how quickly they work.
Earlier, specialists from the Yaroslav the Wise Novgorod State University created a unique mathematical model that allows dynamically to reconfigure hardware neural networks in real time.
Related materials:
AI from Skoltech learned to predict forest fires with an accuracy of up to 87%