The invention discloses a DNS server software fingerprint recognition system and method based on machine learning. The system is composed of an input module, a feature extraction module (for extracting features from search/response), a decision-making tree classification module which is generated through local pre-training, and an output module. According to the system and the method, a target DNS server domain name needing to be detected by a user is taken as input; a program is operated on the background by the system, thereby obtaining a request/packet return feature records of the target DNS server; recognition and judgment are carried out in a trained prediction model; and finally software version information of the target DNS server is output.