The era of voice-controlled devices (VCDs) — such as Amazon Alexa and Google Home — has spawned a generation of smart devices, automated home appliances and cutting-edge cars. But there are problems with this advancement, particularly as regards voice-activated services such as chatbots and attacks on VCDs through audio replay. Our detailed examination of the VCD threat resulted in the discovery that replays may be injected by an attacker in order to break into restricted access to Internet of Things (IoT) devices and nodes. To address this pressing threat, we require reliable, low computation replay attack detection and prevention methods for VCDs and other voice- activated device. In this work, we propose two group delay function domain-based features named Group Delay-Mel Frequency Cepstral Coefficients (GD-MFCC) and Group Delay-Linear Frequency Cepstral Coefficients (GD-LFCC) for feature extraction. Keras calculates the output using pre-trained deep learning models such as CNN or LSTM for the backend processing. The proposed method is tested on Voice Spoofing Detection Corpus (VSDC) to demonstrate its effectiveness.