Android malicious software detection feature extraction method based on deep reinforcement learning

• Main Authors: LI MEIJIN, ZENG QI, YANG TAO, WU YINWEI, CHENG LUYU, FANG ZHIYANG
• Format: Patent
• Language: Chinese; English
• Published: 13.05.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses an Android malicious software detection feature extraction method based on deep reinforcement learning, which relates to the technical field of software and information system security and comprises a sample acquisition step, a deep reinforcement learning model construction step and a model training step. The method is used for carrying out dimensionality reduction on input features when Android malicious software is detected by using a machine learning method, creating an environment and constructing an intelligent agent by using a Double Deep Q-learning Network algorithm, continuously inputting selected features into an Android malicious software classifier to obtain detection accuracy as feedback in the interaction process of the intelligent agent and the environment, gradually optimizing a feature selection strategy, and carrying out feature selection according to the detection accuracy. And finally, redundant and irrelevant features are removed from the originally extracted androi