Rapidly growing cyber infrastructure has reduced the cost of communication and internet therefore, normal people and attackers usages it much frequently. Additionally cheaters are changing their faces therefore secure communication is essential. Among a number of security challenges phishing is one of the issues where a fixed solution is not available. Therefore identification of changing faces of phishing is needed to be investigated. In this work the current techniques based on ML (machine learning) are investigated and categorize methods according to the features used and classifiers. Using the observed techniques a model based ML is proposed. The basic design of the required phishing detection model is reported. Additionally its functional aspects are discussed. Finally the summary is provided and future work is proposed.