Abstract: Methods, devices and systems for detecting suspicious or performance-degrading mobile device behaviors intelligently, dynamically, and/or adaptively determine computing device behaviors that are to be observed, the number of behaviors that are to be observed, and the level of detail or granularity at which the mobile device behaviors are to be observed. The various aspects efficiently identify suspicious or performance-degrading mobile device behaviors without requiring an excessive amount of processing, memory, or energy resources.

Abstract: Methods, systems and devices for generating data models in a communication system may include applying machine learning techniques to generate a first family of classifier models using a boosted decision tree to describe a corpus of behavior vectors. Such behavior vectors may be used to compute a weight value for one or more nodes of the boosted decision tree. Classifier models factors having a high probably of determining whether a mobile device behavior is benign or not benign based on the computed weight values may be identified. Computing weight values for boosted decision tree nodes may include computing an exclusive answer ratio for generated boosted decision tree nodes. The identified factors may be applied to the corpus of behavior vectors to generate a second family of classifier models identifying fewer factors and data points relevant for enabling the mobile device to determine whether a behavior is benign or not benign.

Abstract: Methods, devices and systems for detecting suspicious or performance-degrading mobile device behaviors intelligently, dynamically, and/or adaptively determine computing device behaviors that are to be observed, the number of behaviors that are to be observed, and the level of detail or granularity at which the mobile device behaviors are to be observed. The various aspects efficiently identify suspicious or performance-degrading mobile device behaviors without requiring an excessive amount of processing, memory, or energy resources.

Abstract: Methods, systems and devices for generating data models in a client-cloud communication system may include applying machine learning techniques to generate a first family of classifier models that describe a cloud corpus of behavior vectors. Such vectors may be analyzed to identify factors in the first family of classifier models that have the highest probably of enabling a mobile device to conclusively determine whether a mobile device behavior is malicious or benign. Based on this analysis, a a second family of classifier models may be generated that identify significantly fewer factors and data points as being relevant for enabling the mobile device to conclusively determine whether the mobile device behavior is malicious or benign based on the determined factors. A mobile device classifier module based on the second family of classifier models may be generated and made available for download by mobile devices, including devices contributing behavior vectors.

Abstract: Methods, systems and devices for generating data models in a communication system may include applying machine learning techniques to generate a first family of classifier models using a boosted decision tree to describe a corpus of behavior vectors. Such behavior vectors may be used to compute a weight value for one or more nodes of the boosted decision tree. Classifier models factors having a high probably of determining whether a mobile device behavior is benign or not benign based on the computed weight values may be identified. Computing weight values for boosted decision tree nodes may include computing an exclusive answer ratio for generated boosted decision tree nodes. The identified factors may be applied to the corpus of behavior vectors to generate a second family of classifier models identifying fewer factors and data points relevant for enabling the mobile device to determine whether a behavior is benign or not benign.