Search Results

Spam And Phishing Emails Are Very Troublesome Problems For Mailbox Users. Many Enterprises, Departments And Individuals Are Harmed By Them. Moreover, The Senders Of These Malicious Emails Are In A Hidden Position And Occupy An Initiative Position. The Existing Mailbox Services Can Only Filter And Shield Some Malicious Mails, Which Is Difficult To Reverse The Disadvantage Of Users. To Solve These Problems, We Propose A Secure Mail System Using K-nearest Neighbor(KNN) Algorithm And Improved Long Short-term Memory(LSTM) Algorithm(Bi-LSTM-Attention Algorithm). KNN Classifier Can Effectively Distinguish Normal Emails, Spam And Phishing Emails, And Has A High Accuracy. Bi-LSTM-Attention Classifier Classifies Phishing Emails According To The Similarity Of The Malicious Mail Text From The Same Attacker To Some Extent. By Classifying And Identifying The Source Of Malicious Emails, We Can Grasp The Characteristics Of The Attacker, Provide Materials For Further Research, And Improve The Passive Status Of Users. Experiments Show That The Classification Results Of Attack Sources Reach 90%, Which Indicate The Value Of Further Research And Promotion.