Distribution system protection against faults is a vital task for reliability of a power system. In the proposed work, three phase current signals obtained at substation of an IEEE 13 bus system are initially decomposed using empirical mode decomposition (EMD). The higher frequency components, Intrinsic mode functions (IMFs) of this decomposition are eliminated leaving behind a residue containing predominantly fundamental frequency. This residue being monotonic in nature, is utilized to generate visually distinct symmetrical dot patterns (SDP) for three phases. These SDPs are then correlated with those obtained under normal conditions to obtain a fault index based on alienation method. The faults are successfully detected and classified when the fault index is above a predetermined threshold. This algorithm has been successfully established for various types of faults with changing fault incidence angle, fault resistance at all the buses of the distribution system. The proposed algorithm is found to be fast in detecting and classifying faults within half cycle. © 2019 IEEE.