This article is from the public number:CV Intelligence (ID:CV AI2019) , author: Yu Yang Yang, editor: Zhang Lijuan, plans to title From

In general, there are three common ways to use AI technology for fraud:

The first, sound synthesis, is also the most frequent AI fraud method. The scammer extracts someone’s voice by harassing the phone, etc., and synthesizes the material, and uses the fake voice to implement the fraud.

Second: AI changes face, the credibility of video calls is significantly higher than voice and phone, using AI to change faces, scammers can pretend to be anyone.

The third type: screening the deceived people through AI technology to obtain the chat habits and life characteristics of the deceived. By analyzing various types of information published by the public on the Internet, the scammer will select the target group according to the scams to be performed. For example, when implementing emotional fraud, you can screen out people who often publish emotional information; when implementing financial fraud, you can filter out people who often collect investment information.

In most cases, the AI ​​fraud that people encounter is nothing more than a combination of these three forms. Compared to traditional cyber fraud, telecom fraud, there is no novelty in the situation of AI fraud.

But it is the old-fashioned low-level scam method of telephone, WeChat, and video. Once you wear the AI ​​coat, you can use algorithms to accurately screen specific scam targets, analyze people’s online habits, and grasp people’s psychological weaknesses; Secondly, by changing the sound, changing the face and other means to achieve the effect of false realism, breaking through people’s trust defense, people will easily fall into the “psychedelic trap” created by the scammer.

When AI is used for fraud, who is responsible?

It’s been a while since the first batch of AI scams, and as more and more easy-to-use open source morphing software emerges, the cost of AI scams is getting lower and lower. Not only did the new fraudulent approach not be well governed, but it became more and more rampant.

Google’s WaveNet, Lyrebird speech synthesis software, Adobe’s Project VoCo, and Baidu’s Deep Voice, Keda Xunfei, Tencent, etc. all provide open source applications for speech synthesis.

A court police officer told CV that there is no closed case in China that uses artificial intelligence to defraud. It is common to use AI language imitation software to pretend to be superior.