The multimodal security method is a key method in biometric security systems. Where the system robustness is improved by the integration of different biometric inputs together using different learning methods. Convolution neural network (CNN) approaches were used for deep learning in providing new security measures. The interface of different biometric inputs improves the reliability but increases the complexity and creates a large overhead for the processing system. In this work, the issue of processing biometric inputs with low overhead while preserving the feature details is addressed. A new interface unit is proposed to filter out the relative content for processing. This processing unit offers the advantage of minimizing data overhead and security enhancement using accurate data representation. In this work, energy mapping and low-complex morphological coding is proposed. The proposed work illustrates a higher retrieval performance when compared to CNN based security system. The system is tested on a SDUMLA- HMT dataset for fingerprint and iris image data. The performance of model is evaluated by estimating accuracy, ROC, Sensitivity and processing overhead.