Steganalysis, a field devoted to detecting concealed information in various forms of digital media, including text, images, audio, and video files, has evolved significantly over time. This evolution aims to improve the accuracy of revealing potential hidden data. Traditional machine learning approaches, such as support vector machines (SVM) and ensemble classifiers (ECs), were previously employed in steganalysis. However, they demonstrated ineffective against contemporary and prevalent steganographic methods. The field of steganalysis has experienced noteworthy advancements by transitioning from traditional machine learning methods to deep learning techniques, resulting in superior outcomes. More specifically, deep learning-based steganalysis approaches exhibit rapid detection of steganographic payloads and demonstrate remarkable accuracy and efficiency across a spectrum of modern steganographic algorithms. This paper is dedicated to investigating recent developments in deep learning-based steganalysis schemes, exploring their evolution, and conducting a thorough analysis of the techniques incorporated in these schemes. Furthermore, the research aims to delve into the current trends in steganalysis, explicitly focusing on digital image steganography. By examining the latest techniques and methodologies, this work contributes to an enhanced understanding of the evolving landscape of steganalysis, shedding light on the advancements achieved through deep learning-based approaches.