Many user- and article-level interventions have been proposed to combat the dissemination of misinformation over social networks, including inoculation (exposing users to weaker forms of misinformation), shifting user attention to accuracy, publicizing consensus, fact-checking, using source credibility (while controlling for political polarization), and flagging or providing warnings about an article. Cost pressures and user-engagement risks typically preclude the widespread application of such interventions and motivate a constrained resource allocation-based approach. Restless bandits have been used to model such allocation problems; however, prior work assumes that individual arms only benefit if they receive the resource directly. We note that misinformation-related intervention allocation tasks occur within communities and may be characterized by externalities that allow arms to derive partial benefit when their neighbor(s) receive the resource. We introduce networked restless bandits, a novel multi-armed bandit setting in which arms are both restless and embedded within a directed graph. We then present Greta, a graph-aware, Whittle index-based heuristic algorithm that can be used to efficiently construct a constrained reward-maximizing action vector at each timestep.

In today's world, the spread of disinformation and propaganda online is more widespread than ever. Most of the publisher's revenue comes from advertising, therefore placing ads on these web pages directly funds the publisher, which has been brought under scrutiny by various media. The question of how to remove the publishers from advertising inventory has long been ignored, despite the negative consequences on the open internet. In this work, we make the first step to quickly detect and red-flag the publishers that potentially manipulate the public with disinformation or falsehoods. We build a machine learning model based on multilingual text embeddings that first detects the topic of interest and then estimates the likelihood of the page being malicious. Our systems empower internal teams to proactively, rather than defensively, blacklist unsafe content.

Crowdsourced reporting technologies rely on the public to provide critical information for public decision-making. This work examines security, privacy, and obfuscation in the context of reporting technologies. We show that widespread use of reporting platforms comes with unique security and privacy implications, and introduce a parallel threat model and corresponding taxonomy to outline some of the many attack vectors in this space. We then perform an empirical analysis of a dataset of call logs from a controversial, real-world reporting hotline and identify coordinated obfuscation strategies that are intended to hinder the platform’s legitimacy. We propose a variety of statistical measures to quantify the strength of this obfuscation strategy with respect to the structural and semantic characteristics of the reporting attacks in our dataset.