Face biometric identification has become a preferred method for secure, contactless user authentication. However, the increasing sophistication of spoofing attacks, including photo presentations, video replays, and 3D masks, poses a significant threat to the reliability of such systems. This study presents a novel face liveness detection methodology to mitigate these threats using handcrafted features. The proposed framework utilizes fragmental coefficients derived from transformed face images using Discrete Cosine Transform (DCT), Haar Transform, and a hybrid. These coefficients serve as inputs to various machine learning classifiers and their ensembles. To assess the effectiveness of the method, several performance metrics are used, including accuracy, Normal Presentation Classification Error Rate (NPCER), Attack Presentation Classification Error Rate (APCER), and Average Classification Error Rate (ACER). Experimental evaluations are conducted on publicly available datasets, NUAA, SiW-Mv2, and Replay-Attack. Using 4 × 4 fragmental coefficients from hybrid-transformed face images, the Random Forest classifier achieved 99.99% accuracy in detecting spoof attempts.