Feature Engineering and Computer Vision for Cybersecurity: A Brief State-of-the-Art

1. Related Investigations

Machine Learning (ML) and Deep Learning (DL) techniques have been extensively explored and applied in the domain of cybersecurity to address a wide array of challenges. Notably, (Jha, 2023) surveyed the potential of integrating ML and Natural Language Processing (NLP) for threat analysis and anomaly detection in Smart Grid Technology. Considering the same use case, (Hasan, 2024) explored ML and DL for threat detection in Smart Grid. (Soman, 2023) reviewed thoroughly deep learning and machine learning architectures with their mathematical background such as Naïve Bayes, Random Forest, Deep Autoencoder (DAE), and Deep Neural Network (DNN). (de Azambuja, 2023) works extensively highlighted the ability of ML/DL algorithms such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to enhance the detection and prevention of cybersecurity systems. However, there is a growing need to develop ML/DL techniques capable of handling streaming and real-time data for dynamic threat detection and response in highly dynamic and interconnected environments of Computer Vision. Computer Vision offers novel approaches for threat detection, surveillance, and anomaly identification which (Zhao, 2021) (Ranka, 2023) (Liu, 2023) works overviewed. They informed about the ability to augment traditional cybersecurity measures with visual intelligence capabilities where advanced systems based on CV can analyze visual data from surveillance cameras, satellite imagery, and digital sensors to detect suspicious activities, identify potential threats, and monitor critical infrastructure in real-time. Although CV-based cybersecurity solutions offer a non-intrusive and scalable approach, there is a growing emphasis on the development of multimodal CV systems.