FOCI’21, August 27, 2021, Virtual Event, USA Padmanabhan et al. Figure 1: Timeline of censorship events in Myanmar. Providers (ISPs), like Myanmar, the extent of government-control over the Internet could serve as one indicator of political control, which is difficult to observe in non-democracies [24]. • Collateral damage: Internet censorship in Myanmar showed signs of collateral damage not restricted solely to the country; a notable hijack of Twitter’s address space led to users in other countries being unable to access Twitter as well. We also observe signs of IP address blocking, which has been known to be an aggressive blocking method that could possibly lead to collateral damage, since many domains may be hosted on the same IP address [11, 69, 85]. Indeed, we observe two cases of popular CDN IP addresses blocked that render several sites inaccessible. the President Win Myint, and State Counsellor Aung San Suu Kyi. The coup has been followed by widespread protests as well as violent repression. One might expect the diversity of ISPs in Myanmar today to thwart attempts at nation-wide censorship, but our results show that the military has been able to censor its users surprisingly effectively. Since the coup, there has been an acute increase in Internet censorship in Myanmar, along multiple dimensions. In this study, we examine these censorship events using diverse datasets and show, where possible, their evolution over time. 2.2 Datasets Internet censorship can take many forms—from blanket disruption of users’ Internet connections to targeted blocking of specific domains and keywords. Obtaining a comprehensive view of censorship requires the use of complementary methods that have been designed to detect specific forms of censorship. In this study, we use the following datasets: 2 BACKGROUND AND DATASETS 2.1 Political context and prior censorship Myanmar was ruled by an autocratic military government for nearly 40 years, from 1962 to 2011. In 2007, it became the second country after Nepal to experience government-mandated country-wide Internet connectivity shutdowns [77]. At that time, only two ISPs offered Internet services in Myanmar, both state-controlled [77], facilitating the ability of the government to exert sweeping control. Myanmar saw the end of formal military rule in March 2011, which coincided with decreased Internet censorship [67]. A report published by OONI in 2017 analyzed all OONI measurements collected from Myanmar between Oct. 2016 to Feb. 2017 and did not confirm any cases of Internet censorship, while confirming the accessibility of major social media platforms [81]. Democratic Myanmar experienced a marked increase in the number of ISPs, including international players, such as Telenor and Ooredoo [48]. In June 2019, censorship began to rise again. Internet connectivity was severed in several townships in the Rakhine and Chin states [29, 33, 74] for several months, although such measures have received push-back from service providers [29]. In 2020, an OONI report showed that the scale of Internet censorship had increased, as OONI data confirmed the DNS based blocking of 174 domains on Telenor Myanmar (AS133385). While most of these blocked domains contain adult content, several are ethnic media websites [45]. In Nov. 2020, Myanmar held general elections in which the incumbent National League for Democracy party won in a landslide [51]. On Feb. 1, 2021, a military junta led by Min Aung Hlaing seized control of the government, arresting members of Parliament, • IODA The IODA (Internet Outage Detection and Analysis) system monitors the Internet continuously to identify macroscopic Internet outages affecting the edge of the network, i.e., significantly impacting a network operator (AS) or a large fraction of a country. It uses three orthogonal data sources (Active Probing, BGP, and Internet pollution traffic reaching a darknet) to detect outages and enables visualizing Internet connectivity in nearrealtime on a public site since 2016 [10]. Appendix A.1 provides additional details about these data sources. • OONI The Open Observatory of Network Interference (OONI) [61] project develops free and open source software (called OONI Probe [57]) designed to ethically measure the blocking of websites, instant messaging apps, and circumvention tools. OONI Probe is run by volunteers that have provided informed consent in around 200 countries and territories every month, contributing millions of network measurements from local vantage points. All OONI Probe measurements are automatically submitted to OONI servers, processed, and openly published in near real-time. Since 2012, OONI has openly published more than 420 million measurements from 22 thousand unique AS networks in 239 countries and territories [56]. • Traffic We used a proprietary dataset consisting of aggregated traffic statistics based on NetFlow [41] logs from a large network observability company (Kentik) to analyze user traffic in 28