Phishing is one of the major cyber threats now, where the victims' credentials are obtained by an illegitimate website. This paper proposes a system which will detect old as well as newly generated phishing URLs that have completely no past behaviours to judge upon, using Data Mining. A cloud-based classification model will be created for the same wherein various extracted attributes through the URL will be used as input data. The model will be trained with an exhaustive dataset so as to ensure maximum accuracy.