Censorship Addendum for Advertising-based Measurement: A Platform of 7 Billion Mobile Devices

Given the controversial nature of studying censorship, we had to remove the following information from our submission to MobiCom [4]. We received negative feedback from the MobiSys program committee on this issue and felt that it was not worth the risk in having it rejected at MobiCom. We disagree with the conclusion reached by the MobiSys TPC and in particular we do not believe that the non-interrogative nature of TPC is the best place to evaluate human subjects concerns. You are welcome to cite this work separately from the MobiCom paper. Before conducting the research described here, we sought and obtained the consent of our own Institutional Review Board (IRB) under protocol numbers 20163112 and 2016-3141. Before starting the censorship study we sought additional IRB approval, as similar measurements taken from a website have been considered an area of ethical concern by some in our community [5]. In our study of censorship, users unwittingly retrieve blocked images. Burnett et al[3] used a similar technique of background image loading to measure censorship, but relied on using their own webpage. Their project was submitted to the Georgia Tech and Princeton IRBs, which declined to review the project because it was not collecting PII. Our IRB reviewed our project and has also allowed the study, with restrictions that include using what the Encore paper calls “ubiquitous, yet uninteresting URLs”[3]. These URLs are often found embedded in webpages and constitute non-controversial content (e.g., the Google logo), even if they come from sites generally blocked by governments, such as Facebook and Google; see Fig. 1.