The invention provides a sensitive image recognizing method in interaction of inner and outer power networks. The method includes the steps that (1) sample images are collected in an inner power network system to form a sample image set, and sensitive and insensitive marking is carried out on the sample images in a manual marking mode; (2) feature items of the sample images are selected; (3) a feature data set of the sample image set is extracted according to the selected feature items; (4) a machine learning method is adopted, and a classification model is obtained through training according to the feature data set of the sample image set and the corresponding sensitive or insensitive marks; (5) on the basis of the classification model, sensitive image recognizing is carried out; when the misjudgment rate is smaller than a set misjudgment threshold value, the result that the current classification model meets requirements of an expected target can be determined, and training is finished; when the misjudgment rate is larger than or equal to the misjudgment threshold value, the feature items of the sample images are selected again, and the step (3) is carried out. According to the method, sensitive images are recognized based on the machine learning method, and the excellent classification model can be obtained under the condition that the number of the samples is limited.