Many illegal behaviours could result from a facial recognition and authentication system failure. Systems used for facial recognition nowadays are susceptible to various biometric assaults. The subject of this study is the detection of morphing attacks. This study suggests a reliable detection method that can account for age, lighting, ocular, and headgear variations. A classifier and feature extractor based on deep learning are both used. To improve the detection outcomes, image enhancement and feature combination are also suggested. Moreover, Morph-3 images have not before been discussed in the literature. A more realistic morph attack scenario is shown by professional morphing software. EfficientNet B5 is a powerful deep learning model known for its efficiency and high performance in various computer vision tasks, making it an excellent choice for morph image detection. The system leverages the capabilities of EfficientNet B5 to extract and analyse image features, allowing for the detection of subtle, inconsistencies and alterations in images, a common characteristic of morphed images. The system follows a two-fold approach. First, it preprocesses the input images to enhance their quality and standardize the format. Next, it employs EfficientNet B5 for feature extraction. The deep learning model is fine-tuned on a comprehensive dataset of morphed and authentic images, enabling it to learn the complex patterns and discrepancies associated with image manipulation.