作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2023, Vol. 49 ›› Issue (1): 279-286,294. doi: 10.19678/j.issn.1000-3428.0063413

• 开发研究与工程应用 • 上一篇    下一篇

基于用户行为的社交网络人格特质识别方法

谢柏林, 黎琦, 魏娜, 邝建   

  1. 广东外语外贸大学 信息科学与技术学院, 广州 510006
  • 收稿日期:2021-12-01 修回日期:2022-03-05 发布日期:2023-01-06
  • 作者简介:谢柏林(1982-),男,副教授、博士,主研方向为网络安全、社交网络;黎琦、魏娜,硕士研究生;邝建,讲师、博士。
  • 基金资助:
    广东省基础与应用基础研究基金(2018A0303130045);广州市科技计划项目(201904010334)。

Personality Trait Identification Method Based on User Behavior on Social Network

XIE Bailin, LI Qi, WEI Na, KUANG Jian   

  1. School of Information Science and Technology, Guangdong University of Foreign Studies, Guangzhou 510006, China
  • Received:2021-12-01 Revised:2022-03-05 Published:2023-01-06

摘要: 社交网络已成为人们获取和发布信息的一个重要平台,也是黑客发起网络诈骗的主要场地。大多数黑客在发起网络诈骗之前,首先会判别目标用户的主要人格特点,然后根据主要人格特点制定与其接触的策略。因此,面向社交网络用户的人格特质识别方法的研究对提高用户识别社交网络诈骗能力具有重要意义。提出基于用户的人格特质识别方法。通过构建面向社交网络的人格特质词典提取用户发表或转发文本信息中能反映用户主要人格特质类型的观测值,采用5个具有不同参数值的隐半马尔可夫模型刻画用户在社交网络上发表或转发文本信息的行为过程。在人格特质识别阶段,通过计算每个用户在发表或转发文本信息过程中产生的观测序列相对于模型的平均对数似然概率,以识别用户所属的人格特质类型。在采集的新浪微博数据集上进行实验,结果表明,当假正率为10%时,该方法的总真正率为93.18%,能准确识别用户的人格特质类型。

关键词: 社交网络, 人格特质, 隐半马尔可夫模型, 用户行为, 网络诈骗

Abstract: Social networks have become an important platform for people to obtain and release information, and are also the preferred sites for hackers to launch an online fraudulent scheme.Before starting such a scheme, most hackers will first identify the main personality characteristics of the target users and then formulate contact strategies on the basis of these characteristics.Therefore, studies on personality trait identification methods for social network users are of great significance to improve users' ability to identify fraudulent schemes on social network.In this paper, a user-based personality trait recognition method is proposed.By constructing a personality trait dictionary for social networks, the proposed method can extract the observation values that can reflect the main personality trait types of users based on the text information they have published or forwarded.Moreover, the method uses five Hidden semi-Markov Models(HsMM) with different parameter values to describe the behavior process of users when publishing or forwarding text information on social networks.In the personality trait recognition stage, the average log likelihood probability of the observation sequence generated by each user while publishing or forwarding text information relative to the model is calculated to identify the type of personality trait the user has.The experimental results on the collected Sina Weibo dataset show that, when the false positive rate is 10%, the total true positive rate of the proposed method is 93.18%, which indicates that it can accurately identify the user's personality traits.

Key words: social network, personality trait, Hidden semi-Markov Model(HsMM), user behavior, Internet fraud

中图分类号: