Abstract : This work analyzes the complexity and potential of secure adaptive filter in adversarial environment and presents a set of secure adaptive filter algorithms (SAFA) based on statistical learning theory, self non-self binary classification and also multi-class classification. The basic building blocks of secure adaptive filter algorithms are association rule based classifier, decision tree and optimal margin classifier. Secondly, SAFA verify the collective and security intelligence of the filter adaptively in adversarial environment in terms of goals, inputs, outputs, mechanism, reliability and consistency of system performance, correctness, fairness and rationality of rules configuration and risk of corruption and various types of malicious attacks. The scope of adaptive filters has been analyzed in terms of a set of interesting database and communication networking applications, distributed systems, data streaming, packet classification, spam filtering, web caching and may be extended to the other application domains such as cyber security, defense, broadcast and mobile communication, finance and governance. Specific focus has been given to correctness (e.g. false positive and false negative) and privacy of computation by the adaptive filter. The correctness of the filter is deeply associated with right configuration of the data schema. The privacy of computation can be ensured through a robust revelation principle of the filter.
Keywords : Secure adative filter, Threat analytics, Algorithms, Rule based classification, Decision Tree, Optimal margin classifier, Communication, Cyber security