The invention discloses an authentication method based on application login, and the method comprises the steps: extracting the historical login data of a user, and creating a behavior prediction regression model; according to the behavior prediction regression model, identifying a new access request of the user, performing matching to obtain an optimal authentication mode and an access path, andfeeding back the authentication mode to the user; and obtaining authentication information fed back by the user, and opening a corresponding page according to the predicted access path. According to the invention, the login data of the user is extracted; big data storage is utilized, the data is analyzed and trained through machine learning to obtain the customized regression model, and the most common and most convenient login authentication mode of the user is intelligently predicted according to the login history and habit of the user, so that the interaction steps between the user and theapplication system are reduced, the operation efficiency of the user is improved, and the user experience is improved.