Method for detecting image-based spam by utilizing image local invariant feature

Abstract

The invention relates to a method for detecting image-based spam by utilizing the image local invariant feature, which comprises the steps of extracting invariant region feature of spam information in an image by utilizing the accelerated extraction algorithm with the robust feature, further generating a feature vector of the image, estimating parameters of a Gaussian mixture model by using the maximum likelihood algorithm and training a classifier of the Gaussian mixture model. Experiments show that the method can improve the recall rate of the spam and save the program computation time and the space. The classifier based on the Gaussian mixture model is obtained. The realizing method for detecting the image-based spam comprises three modules of the extraction of the image feature, the estimation of the parameters of the Gaussian mixture model and the detection of the image-based spam.

Description

technical field [0001] The present invention is a realization scheme of using the local invariant features of garbage pictures to train a Gaussian mixture model to detect image-type spam, which mainly solves the problems of low detection efficiency and recall rate of picture-type spam in today's technology, It belongs to the field of data mining and machine learning. Background technique [0002] E-mail has become an important way for people to communicate on the Internet, but due to the huge commercial, economic and political interests, the amount of spam has increased dramatically. The spam that was prevalent at first was to embed spam information such as advertisements into images in the form of text. Hrishikesh et al. are using the mined text and color features to classify emails [1]. In 2006, Fumera et al. proposed an OCR (Optical Character Recognition) technology to detect the text information of image spam, which has a better detection effect than other filtering sys...

AI Technical Summary

Problems solved by technology

In 2008, Mehta et al. detected a large number of spam generated by templates. Using repeated similarities, the accuracy of SVM classification reached 98%. At the same time, they proposed an algorithm for clustering images using GMM [6]: Each picture is reduced to 100×100 pixels, the texture shape and color features of each pixel are extracted, GMM is trained for each picture, and the similar distance in the picture is calculated to cluster, and the garbage picture is calculated by calculating the threshold, although This method is accurate but the amount of calculation is too large, and the time complexity of the algorithm is high, which is not conducive to practical application.

Images