Research And Implementation Of Social Engineering Attack Defense System Based On Dialogue Agent

With the rapid development of the Internet and the increasing social needs of people,social engineering attack has become one of the most popular attacks in the network.This kind of attack is different from the traditional network attack,which is not limited by the firewall,intrusion detection mechanism or other computer defense software,directly use the victim's greed,curiosity and other psychological weaknesses to cheat,resulting in property loss,information disclosure and some serious consequences.Therefore,it is of great significance to establish a perfect social engineering attack defense system.This paper proposes a social engineering attack defense method based on dialogue agent for real-time interaction scenarios such as social applications and social websites.This method integrates the functions of attack detection,attack blocking and automatic reply.After detecting the social engineering attack,it completes the interaction process with the attacker on behalf of the user.When blocking the attack,the system selects the appropriate dialogue strategy to capture the attacker's information,so as to make the whole process proactive.The interaction function of dialogue agent is the core of the proposed method.In order to obtain the attacker's information effectively,the paper proposes a multi sentence pattern dialogue generation model,which improves the initiative of dialogue agent by controlling the sentence pattern of the response.This model combines Seq2Seq model with Autoencoder by sharing the decoder,and introduce loss functions such as Euclidean distance and Triplet loss to make semantic variable with different sentence patterns separated in vector space,thus realizing the sentence pattern controllability of the dialogue model.At the same time,latent variable is introduced into the model as additional input of decoder,which helps the model extract sentence information and improve the quality of text quality.The experimental results show that the model can generate responses with designated sentence patterns and achieve a high level of content fluency and diversity.Based on the method,this paper designs and implements a social engineering attack defense system based on the dialogue agent.The system includes user management module,dialogue accessing module,social engineering attack detection module,dialogue strategy management module,text generation module,contact management module and text sending module.The main functions and processes of the modules are described in detail,and the test proves the feasibility and practicability of the system function.The results show that the system can help users defend against social engineering attacks in the process of daily network conversation and provide security for personal information.