Attacker or attackers’ group fabricates fake news with bad influence to launch a public opinion attack against an individual or company by purposefully spreading on major forums and media. Most commonly, the malicious short-sellers attack targeted public companies by spreading rumors and disseminating distorted articles, resulting in stock price fluctuations or even sharp falls, so as to make profits therefrom. Unfortunately, tracking attackers is challenging because most rumors and fake news are published anonymously or through pseudonyms.
Through the analysis of a large number of articles, it’s found that different articles of the same author have similar content and stylistic features. We propose a novel deep learning model, which learns a mapping from articles to a compact Euclidean space where distances directly correspond to a measure of content and stylistic features’ similarity, so as to find out the real identity of the attacker. We construct a data set containing 100000 articles for the experiment, and our model can obtain significantly better accuracy at identifying the new authors outside the training dataset.