Bad Ai: Investigating the Effect of Half-Toning Techniques on Unwanted Face Detection Systems

摘要

Currently, automatic face detection systems are widely used in various social media. The improvement of the automated intelligent face recognition can easily result in the invasion of the user 's privacy if the predefined functions for their so- called artificial intelligent systems are designed with no respect to the user privacy, this means that users do not get the right to prevent these types of automatic face detection. The problem does not stop here, and recently some organization attempted to use automated face recognition system as a digital personal judgment tools. For instance, judging if a person is criminal only based on the attributes of his/her face. In the present technology, spreading the personal information such as individual digital images are fast and inevitable. Thus, in this paper, some of the techniques to evade automatic face recognition is investigated. The focused of this article is to highlight the advantages of different Half-Toning algorithms concerning avoiding unwanted automated face detection/recognition. In the experimental phase, the result of varying filtering and modification algorithms compared to the proposed filtering technique.