Main Article Content
Abstract
Integrity violation, availability violation and privacy violation are the major intrusion caused in the network environment. An integrity violation, if it allows the adversary to access the service or resource protected by the classifier. An availability violation, if it denies legitimate users access to it. A privacy violation, if it allows the adversary to obtain confidential information from the classifier. Pattern classification techniques are extended with adversarial settings. Secure pattern classifiers are designed to control the performance degradation under potential attacks. Secure pattern classifier framework is build with model selection and training and testing data construction method. TR and TS construction algorithm is used to select data for pattern classifier. Secure pattern classifier is improved with attack control mechanism to handle the intruders in the testing process. Training patterns are secured with simulated attack patterns. Pattern update process is monitored and controlled in testing process. Classifier utility rate is improved in testing process.