Unconstrained Facial Recognition (FR) entails a complex version of traditional face recognition problem involving challenging and varying conditions. The existing research focuses on one of the core challenges in facial recognition and does not account multiple features for recognition. This paper presents an Unconstrained Multi-Feature Facial Recognition (UMF-FR) system for robust and secure unconstrained facial recognition. The UMF-FR system performs discriminative feature extraction relying on multiple feature channels including Gabor features, ResNet features, Inception features, and VGG features. These multiple features form the basis for fused feature vector for final classification using deep convolutional neural network (CNN) architecture. The UMF-FR model is trained over multiple hyper-parameter settings on LFW dataset. The results indicate promising performance of UMF-FR approach and establish a platform toward solution of facial recognition in unconstrained conditions.