Online communication is a fairly recent phenomena that our species has yet to be adapted to. It lacks several native communication media — such as tone of voice and body gestures — which our species has become dependant on to understand day-to-day communication. Subsequently, users may have difficulties just in understanding each others even in benign circumstances. The unnatural communication media also makes judgement of online deception difficult. Often this deception is targeted at large crowds with the intent of deceiving a massive number users simultaneously, such as in spam emails and product review comments.
Recently, automated text generation has become a real threat in online communication. A malicious user may use deep neural networks to automate the generation of texts with the intent to deceive people. In some contexts like restaurant reviews, generated text is now natural enough to fool human readers into thinking that the messages have been written by actual people. We believe machine-aided detection of fake reviews is necessary now, since humans lack the ability to differentiate real and the new generation of machine-generated reviews.
Publications
Mika Juuti, Bo Sun, Tatsuya Mori, N. Asokan. Stay On-Topic: Generating Context-specific Fake Restaurant Reviews. arXiv preprint arXiv:1805.02400