Methods and devices for detecting suspicious or performance-degrading mobile device behaviors may include performing behavior monitoring and analysis operations to intelligently, dynamically, and/or adaptively determine the mobile 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 behaviors are to be observed. Such behavior monitoring and analysis operations may be performed continuously (or near continuously) in a mobile device without consuming an excessive amount of processing, memory, or energy resources of the mobile device by identifying hot application programming interfaces (APIs) and hot action patterns that are invoked or used most frequently by software applications of the mobile device and storing information regarding these hot APIs and hot action patterns separately and more efficiently.