Network blasting attack detection method and system based on langchain large model

• Main Authors: WEI YONG, XU XIAOBO, LUO PENGYU, SUN CHENGSHENG
• Format: Patent
• Language: Chinese; English
• Published: 04.06.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRIC DIGITAL DATA PROCESSING; ELECTRICITY; PHYSICS; TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHICCOMMUNICATION

The invention provides a network blasting attack detection method and system based on a langchain large model. The method comprises the following steps: S1, loading a knowledge base; s2, performing data segmentation on the knowledge base loaded in the step S1 to obtain text blocks; s3, combining the text block obtained in the step S2 with the embedded model, and outputting an information text fragment; s4, constructing a cue word template according to the information text fragments output in the step S3; s5, customizing a network security domain task customization data set, and completing model pre-training; s6, performing fine tuning and evaluation on the pre-trained model; and S7, combining the cue word template constructed in the step S4 with the fine-tuned and evaluated model in the step S6 so as to judge the network blasting attack. According to the method, the blasting attack text can be well mined, the blasting attack information can be found, and the blasting attack behavior in the network flow can be