Method for detecting privacy leak of user based on multi-task learning, server and system

Abstract

The invention discloses a method for detecting privacy leak of a user based on multi-task learning, a server and a system. The method for detecting privacy leak of the user based on multi-task learning comprises the following steps: for comprehensive representation of privacy of the user, dividing the privacy of the user into a plurality of fine-grained privacy categories in advance, and dividingthe fine-grained privacy categories into a plurality of groups to form group structural information for privacy of the user; extracting privacy characteristics of the user from different direction torepresent the privacy categories of the user in an all-round manner; constructing a prediction model based on the extracted privacy characteristics of the user; and introducing multi-task learning, enabling each task in the same group to share related characteristics, and performing characteristic grouping by taking the group structural information of the privacy of the user as a priori by use ofa group lasso model. Therefore, the modeling performance for detection of privacy leak of the user and interpretability of the model are improved.

Description

technical field [0001] The invention belongs to the field of information classification processing, and in particular relates to a user privacy leakage detection method, server and system based on multi-task learning. Background technique [0002] With the rapid development of Web2.0, the Internet has entered the era of social media, and social media has become an important platform for people to obtain and share information on a daily basis. As users are increasingly exposed to social media, privacy threats have gradually become an important issue related to the vital interests of every online user. [0003] In recent years, there have been many research works around the detection of user privacy leaks in social media. Existing techniques are mainly applied to structured data. For unstructured data, effective classifiers are usually trained intensively, but they mainly focus on coarse-grained privacy judgments and ignore the correlation of privacy, making the model less in...

AI Technical Summary

Problems solved by technology

Existing techniques are mainly applied to structured data. For unstructured data, effective classifiers are usually trained intensively, but they mainly focus on coarse-grained privacy judgments, ignoring the correlation of privacy, making the model less interpretable

Images