Toward Hybrid Classical Deep Learning-Quantum Methods for Steganalysis

Abstract

This paper explores the potential of the hybrid classical deep learning quantum in addressing contemporary information security challenges, including cyber-attack detection and prevention, botnet detection, IP-theft, hardware defect detection, and more, with a specific emphasis on steganalysis. We begin by offering a succinct overview of current information security research trends, focusing on the hybrid classical deep learning-quantum paradigm. Subsequently, we introduce a classical-quantum hybrid framework tailored for uncovering concealed information within a significant public steganalysis dataset, ALASKA2. To enable image extraction and classification, we present a two-layer architecture that integrates convolutional neural network layers into both the classical deep learning and quantum layers. Notably, the proposed approach has improved steganalysis accuracy by achieving a 97% implementation accuracy rate.