The uncertainties of energy networks have increased in recent years due to the fast and widespread penetration of renewable resources. In this paper, the energy market trading of green microgrids composed of wind and solar units is taken into consideration under the information vulnerability of renewable energies. Since wind speed and solar radiation data are the most critical parameters to calculate the output power of wind turbines and photovoltaics, it is assumed that non-legitimate agents attempt to alter them and inject false data toward increasing operational costs. In order to mitigate the influence of this problem, a data-driven framework consisting of evaluation, purification and prediction parts is designed in which the k-nearest neighbour algorithm is utilized for anomaly detection and various methods including artificial neural network, deep learning, Gaussian process, linear regression and support vector machine are implemented and compared to specify the best operation for prediction unit. It should be noted that a stochastic approach is also used to model probable malicious attacks and avoid any biased behavior. The results validate that the proposed framework supports the operator to make better decisions for participating in day-ahead and real-time energy markets in the presence of renewable resources vulnerability.