Achieving Differential Privacy And Fairness In Machine Learning