Research On The Application Of Learning To Rank In Personalized Recommendation Systems

The computer network technology and Internet communications are rapidly developing, and the wireless network becomes increasing popular. The most core issue of the information age is network information security. The disclosure of sensitive information is becoming a serious threat to network communication. Most of the eavesdropping over Internet use special network protocols which do not publish their protocol specifications, cannot use the existing preventive measures. In order to protect network environment, to detect and warn the malicious attacks, unknown protocol identification is becoming a significant topic.The unknown protocol researched here should have fixed certain format, rather than lacking of format rules. In this thesis, we intent to infer the unknown protocol’s format and specification from protocol message patterns using semantic mining algorithm and technical scheme. We ultimately provide a technical solution in unknown protocol messages identification in this thesis.The protocol keywords and protocol format generally locate in control- frame. By using unsupervised machine learning K-Means algorithm based on vector space model, the control- frame can be recognized and marked from protocol messages. The extracted control- frame helps improve the efficiency of protocol semantic mining and protocol format identification. Protocol format contains keywords that characterize the protocol function and states. It is the key point of protocol identification. The Latent Dirichlet Allocation topic probability model dealing keyword identification and initiate format exaction is used in this thesis. The initiate format from semantic mining is unordered and redundant. With the Needleman-Wunsch sequence alignment algorithm the protocol format can be extracted. And building the finite-state machine can infer the protocol grammar accuracy.The experimental results show that the K-Means machine learning algorithms and the Latent Dirichlet Allocation semantic mining model can efficiently extract the keywords and format of the unknown protocol, and further infer the protocol grammar. The protocol grammar can identify the network traffic more accuracy. The semantic extracting of the unknown protocol can identify the network traffic and messages accurate which provides a basic technology for attack warning in network communications.