Adversarial Text Generation for Personality Privacy Protection
Zhe Wang, Kangfeng Zheng, Qingbiao Li, Maonan Wang, Xiujuan Wang · 2021
Protecting the user's personality privacy can effectively interfere with or deceive the attacker's personality analysis, avoid the attacker's use of personality vulnerability, and reduce the success rate of social engineering attacks. However, the current research on personality privacy protection is at a blank stage. To solve this problem, we propose a personality privacy protection method based on adversarial text generation. This paper mainly uses gradient-based adversarial method and cosine similarity to generation adversarial text. We formed a set of replacement words to test the impact of the number of replacement words on the performance of the model. Experiments show that the method proposed in this paper has achieved good effects on model attacks (reducing the performance of the model), and can well complete the task of protecting personality privacy.