Design And Implementation Of Filtering System For Security Information Sensitive Words Based On Aggregate Tree Matching

A large number of interactive information will be generated by the various notifications of securities trading and the daily care of customer in the IT system of securities companies.These information are usually pushed in the form of short messages,which needs to be filtered by sensitive words before they are sent to the customers,avoiding the negative impact of the bad SMS(Short Messaging Service)on a large scale.Therefore,a sensitive word filtering method based on aggregate word tree is studied and optimized.This filtering method is composed of two stages,one is the initialization of the sensitive word library,the other is the matching of the sensitive words.In the initialization stage,many word chains are produced by aggregating different words in the sensitive thesaurus,then multiple word chains are aggregated into a word tree and loaded into the cache.The sensitive word matching phase begins after the initialization stage is completed:the target information is read and each word in it is scanned,judging whether the target information matches the word chain in the sensitive word tree,the match hit will be counted in the cache,and the alarm is triggered when the count reaches the threshold.Furthermore,the filtering method is optimized to be used in large-scale SMS processing scene by introducing multi thread processing mechanism in the initialization stage of sensitive words and using cache to manage the aggregate word tree caching.The target messages are also stored in the cache to make the matching phase can be executed in memory,which can improve the filtering performance significantly.On the basis of above study and optimization,a sensitive word filtering system based on aggregate word tree matching is designed,and the system requirement analysis,overall architecture design,data structure design and so on are given.In the end,the implementation and testing of the system are also provided,which indicates that the proposed method of filtering securities information sensitive words can provide a good data security early warning mechanism for the IT system of securities dealers.