Should We Trust Influencers on Social Networks? On Instagram Sponsored Post Analysis

With online social networks (OSNs), people are exposed to tons of fake information or misleading posts. Celebrities sometimes intentionally create misleading posts in OSNs to guide people for commercial or marketing purposes. The intentional phrases from such posts can affect the online rating and even lead to a frenzy shopping. That’s part of the reasons that the top social media influencers are targeted by merchants to help promote products. The Federal Trade Commission (FTC) requires that all sponsored posts must be clearly disclosed. However, many influencers do not follow the FTC rules. As a result, people may be misled by the undisclosed sponsorship. In this study, for the first time, we explore the credibility of posts on Instagram and analyze if an influencer complies with the FTC requirements. We build an effective Undisclosed Sponsored Post Detection (USPD) framework based on an ensemble of machine learning classifiers. The USPD framework consists of three main processes: (i) feature extraction, (ii) model construction and (iii) credibility and integrity analysis. Our analysis and experiments demonstrate that the proposed framework can achieve a high accuracy of 83% for undisclosed sponsored post detection. The proposed framework also takes advantages of the text, user and image features in OSN posts to effectively analyze how much an influencer can be trusted.