From past few years, Automatic Speaker Verification (ASV) systems had improved their accuracy by making popular in practical use for security systems by offering secure authentication method. Traditional approaches for speaker identification and verification had faced several challenges which include speaker mimicry, environmental noise, and computational complexity. Therefore, this research proposes Deep Neural Network-Improved Wild Horse Optimization (DNN-IWHO) for identification and verification of speaker. Initially, data is collected from Voxceleb1dataset and preprocessed by using data augmentation, white noise, and pitch shifting to enhance generalization. Then, feature extraction in done by using spectrum chroma and Mel-Frequency Cepstral Coefficients (MFCC) to represent the feature vectors of a large set of data, and these features are identified and verified by using proposed DNN-IWHO. The proposed DNN-IWHO model acquired higher results with the accuracy of 98.89%, F1-score of 99.44%, precision of 98.77%, recall of 98.56%, when compared with the existing Multi-scale Recurrent Network (MRN).