Sensitive data identification and classification technical method based on image identification

Abstract

The invention discloses a sensitive data identification and classification technical method based on image identification. The sensitive data identification and classification technical method based on image identification comprises a sensitive strategy model construction method and an image sensitive information mark classification method. According to the invention, an OCR (Optical Character Recognition) technology is adopted to directly extract characters from the image; sensitive information classification is carried out on the images according to a preset sensitive information strategy, image information can be recognized and classified more accurately, the sensitive information strategy can be directly modified in the later period, image classification training does not need to be carried out any more, the expansibility is higher, sensitive classification is more detailed, and the service difference requirements of different service systems can be met. The method can adapt to andmeet service difference requirements of different service systems, image text extraction is carried out, sensitive information is rapidly and accurately recognized by using the sensitive strategy, sensitive feature extension is realized, user-defined sensitive information matching of images is realized, and private data leakage is reduced.

Description

technical field [0001] The invention belongs to the field of data security technology and machine learning, and in particular relates to a technical method for identifying and classifying sensitive data based on image recognition. Background technique [0002] Aiming at the technology of sensitive recognition in images, there are currently image classification methods based on deep learning. The core is to assign a label to the image from the set classification set, identify objects, scenes, behaviors and other information in the image, and return the corresponding label information. This method must first set a fixed classification and carry out a lot of training before it can classify pictures. The new classification is too difficult and sensitive information cannot be extracted. [0003] The machine learning algorithm TENSORFLOW (a symbolic mathematical system based on data flow programming) is used to train the system to recognize each object, and the training image data...

AI Technical Summary

Problems solved by technology

Therefore, the existing technology has the following disadvantages: 1. The amount of sample data required for learning is large, and each image type needs to provide a large number of sample images; 2. The amount of calculation is large, the requirements for computer hardware are relatively high, and the machine learning training model takes a long time ;3. Modifying the classification samples requires re-learning; 4. Different learning rates will lead to different results

If the rate is too large, the accuracy rate will continue to jump up and down during the training process. If the rate is too small, the expected accuracy will not be achieved before the end of the training; 5. For images that are similar but have different types of text sensitive information inability to distinguish carefully

[0005] It can be seen that image classification by machine learning algorithm TENSORFLOW cannot satisfy the precise classification and extraction of text information in pictures, so we need to propose a technical method for sensitive data identification and classification based on image recognition