Research Of Artificial Neural Networks On SMS Spam Identification

Gradually into the information age, an obvious sign: With the rapid increase in China’smobile phone users, SMS business has become one of the leading business of the majoroperators，it has Shortcuts and hidden features that make it the new darling of the people tocommunicate and exchange. But it has the convenience of public life at the same time it alsobrings some of the information security aspects, Many criminals use SMS to send ads andsend false SMS fraud, it is affecting the normal life of the people. How to maintain the healthand vitality of the development of the short message service at the same time to avoidinfringement of spam messages on people’s lives that has become a pressing problem. Thereare two main types of solution: From the level of legal and ethical constraints, to appeal to themasses in reporting spam messages and bad numbers; Leaving from the technical level,mainly through the study of Series of filtration systems, Filtering and monitoring spammessages. But by the actual using, there are a lot of deficiencies in these two solutions. To beconstrained only from the law is not enough. In the technology sector, the establishment ofspam messages to monitor the number of all messages sent in the form of the blacklist, andfilter independent adverse and malicious words, there are a lot of blindness and thephenomenon of shielding normal SMS. How more accurate separation, extraction out of thespam messages, we need to study.The identification of spam messages is the non-linear classification problem. From thischaracteristic, this paper applies the knowledge of artificial neural network to solve suchproblems. The main method steps: More serious harm to fraudulent spam messages as aresearch object, extracting the fraud special testimony as input neurons and types of fraud asthe output neurons, using of BP neural network learning algorithm to establish a network oftraining and testing samples; Come to the classification results, and set a good classificationfor comparison and analysis.Experiments showed that the result of the BP network training is the same as manualclassification. Adding confounding factors in the sample, and still obtained good results. Inthe field of recognition processing of spam messages, this paper had some progress and thevalue of practical application.