Detection system of software implants using decoys

The article presents a study on a method for detecting malware that employs steganographic images for covert system control. The proposed approach is based on using neural networks to detect steganographic alterations and identify abnormal software behavior. The research involved a series of experiments aimed at comparing the performance of the proposed neural networks (CNN and LSTM) with traditional methods, such as decision trees and Bayesian classifiers, as well as analyzing the effectiveness of various steganography detection techniques. The experimental results demonstrated that neural networks significantly outperform traditional approaches in terms of sensitivity, specificity, and overall accuracy, achieving 99.2% accuracy in detecting steganographic changes and 98.75% accuracy in malware identification. Despite the high effectiveness, the method has limitations related to the use of separate models for different stages of analysis, which may complicate its application in resource-constrained environments. Future research will focus on optimizing the method by integrating the models into a unified system and enhancing resilience to new types of attacks. The proposed approach has the potential to be integrated with existing information security systems, providing an increased level of cybersecurity and automated threat detection.