The long-run effect of Cuba's recent Internet-augmented protests

Neither tweets and Facebook video clips nor the repetition of tired slogans from Granma are likely to convince many Cubans to change their minds.

It's now more than 6 weeks since the Cuban political protests and accompanying Internet service disruption. Will they lead to a long-run change in the Cuban Internet or the Cuban political situation?

Let's start with the Cuban Internet. Many of the Internet changes during the protests have disappeared. Total daily traffic, the ratios of mobile to fixed traffic, and human to automated posts, and the proportion of blocked Signal sessions are about what they were before the protests.

But everything is not the same. Before the protests, tech-savvy Cubans had been using the Psiphon VPN service to access content the Cuban government blocked. In the ten days leading up to the protests, Psiphon averaged 17,285 unique daily users.  During the protests, daily users peaked at 1.425 million on July 16, then began to taper off. However, it did not return to the pre-protest level. For the 13 days shown here, the average was 109,429 unique users per day, over six times the number during the runup to the protests.

Speed measured at the
University of Havana

Keith McManamen from Psiphon said that this kind of retention -- 5x-10x baseline -- is typical of their experience in other nations and that Cuban usage intensity -- bytes/unique user/day and sessions/user/day -- has returned to pre-protest levels. Since Psyphon access is slow and Internet access is expensive in Cuba, roughly 90,000 new users must be motivated.

Another Internet discontinuity is the existence and continued growth of a crowd-sourced archive of images and videos of the protests. There were 219 documented protests as of August 6 and today there are 281. This archive may inspire future discontent or it may be used to identify and prosecute protest participants. Regardless, it will be available to historians and political scientists -- either at its present URL or, if it's deleted, on the Internet Archive. (As an aside, the first such historical protest archive that I know of contains all of the Usenet traffic during the protests against the 1991 Soviet Coup attempt and it is still available online).

There is also speculation that the protests triggered or accelerated the new Internet regulation, announced on August 17th. The regulation treats online content as a potential security threat and bans “spreading fake news,” “slander that impacts the prestige of the country,” “inciting protests,” “promoting social indiscipline,” and undermining someone’s fame or self-esteem.”  The government has also set up a Web site for citizens to report violations -- crowd-sourcing 1984.

Cuba has had previous protests, but the prolonged, nationwide protests that began on July 11th were a product of mobile Internet access, which began rolling out in 2018. The protests led to the changes to the Internet mentioned above, but will the Internet be the cause of long-run political or social change? 

I suspect not.

 

Cloudflare reports that during the last 30 days, 70% of Cuban traffic was from mobile phones and 30% from fixed desktops, as it was before the protests. Furthermore, the majority of the fixed traffic is from workplaces where access is controlled and easily surveilled. Therefore, most political traffic is to and from mobile phones, which means posts are short and divisive -- government supporters are talking to and reinforcing the beliefs of other government supporters and the same goes for the protesters. (There are many beneficial Internet applications in business, science, health care, education, entertainment, etc., but they generally depend upon the fixed Internet).

I don't know what percent of Cubans support or oppose the government, but neither tweets and Facebook video clips nor the repetition of tired slogans from Granma are likely to convince many Cubans to change their minds. Reasoned, long-form argument on the fixed Internet might be a little more persuasive, but things like money, power, organization, demographics, negotiation and compromise and exogenous factors like climate change, COVID, or political and economic decisions made in China will determine Cuba's political and economic future, not the mobile Internet.

 

The Cuban Internet in the aftermath of the anti-government protests

Before the protests, around 18 thousand Cubans used the Psiphon VPN per day. Now its holding steady at around 166 thousand per day.

In an earlier post, I looked at the use of the Internet by anti-government protesters last month and the government's attempt to block them. Now, a few weeks later, let's see how the Internet changed after my July 18 post.

The protesters used messaging and social media services, which the government tried to block, and posted images and videos of protests around the island. The following is a snapshot of an interactive, crowd-sourced map showing the locations of the demonstrations (200 on the 11th, 11 on the 12th, 5 on the 13th, and 3 on the 17th). There are now 219 documented protests and you can see images and videos by clicking on the locations

The governmenet began blocking access to Web sites and messaging and social media services the evening of the 11th. For example, in my previous post, I noted that OONI, an Internet monitoring organization, reported that the encrypted messaging service, Signal, was blocked at 2:41 AM UTC on July 12th but drilling down on OONI data, we see that the blocking is sporadic, not constant and that must frustrate and discourage users.

As shown here, OONI reported that Signal was heavily blocked during the days of the protests, but no tests were blocked on July 19-21 from the 13th through the 28th.

OONI test results are reported when someone runs a query using the OONI probe application. Some of the variation in results is probably due to tests being run from different locations since, for privacy protection, they do not geo-locate within a nation.

My biggest surprise while looking through OONI data was to discover that the Website of the anti-regime Cuban American National Foundation was accessible on July 26th at 19:24 UTC and July 27th at 15:43 UTC. Can anyone offer an explanation for that?

Cloudflare monitored the Cuban Internet during the protests and noted a marked shift from mobile to desktop traffic. As shown below, that shift has largely been reversed, but a visible inspection of the plots indicates that mobile traffic has not completely recovered to pre-protest levels. 

There was also a sharp increase in the percent of traffic written by bots rather than humans during the protest and, as shown below, humans have more than recovered.

The Cuban government attributed the increase in the percent of automated traffic to the US government and Cuban mercenaries, but an analysis by Cazadores (hunters) of Fake News does not support that assertion. They downloaded and analyzed 1,048,576 tweets with the #SOSCuba hashtag that were posted between July 10 and 12, 2021. Only 4.84% were "pure" tweets, 83.68% were retweets, 8.55% were replies, and 2.83% were "other." The study concluded that

• The contribution of bots and spam was insignificant relative to real, organic tweets from Cuban citizens on the island and supporters -- human rights defenders, journalists, bloggers, and artists -- in other nations.
• The vast majority of the photos and videos that were disseminated on Twitter were authentic. 
• Ninety-three "highly suspicious" accounts published more than 144 tweets per day and generated 27,442 tweets, but that was only 2.66% of the total. (I looked at the most prolific, @RRetuiter, and it is apolitical random nonsense).

As shown below, the overall level of Cuban traffic peaked during the protests and has dropped back off, but it is still noticeably higher than before.

During the protest period, technically oriented Cubans used the Psiphon VPN service to circumvent blockage and to my mind, the most interesting post-protest result is the rate of Psiphon VPN utilization, as shown below.

Before the protests, daily usage was consistently around 18 thousand unique users per day. (It was up to around 28 thousand on the first day of the protests). There were nearly 1.4 million users during the heaviest protest day and it now appears to be holding steady at around 166 thousand. It seems that knowledge of the Psiphon VPN spread during the protest period and people learned to use it. I wonder how many Cubans downloaded the client during the protests.

Subsequent to the protest, CyberGhost, a Romanian VPN service, offered a free account to any Cuban users. It would be interesting to see a Cuban speed comparison between Psiphon and CyberGhost.

The bottom line is that usage patterns have shifted toward pre-protest levels, more Cubans know about VPNs and more people outside of Cuba are aware of the situation there.

The Internet during the anti-government protests in Cuba

On Sunday, July 11, thousands of Cubans, took to the streets in anti-government protests triggered by COVID, the faltering economy and an overwhelmed healthcare system. In four days, at least 118 large and small protests took place across the island. The following is a snapshot of an interactive, crowd-sourced map showing the locations of the demonstrations (94 reported on the 11th, 14 on the 12th, 7 on the 13th and 3 on the 17th).

For the interactive version, many with images and videos, click here. As the images and photos show, the government responded with arrests and violence in some cases.

The government also began blocking the Internet on the evening of the 11th, as shown in the following graph, posted by Doug Madory, Director of Internet Analysis at Kentik.

It was totally blocked for a short time, then partially blocked. Madory speculated that they may have been trying to figure out how to block certain portions of the country.

Several messaging apps were blocked during the protest period. As you see below, the encrypted messaging app, Signal, was blocked at 10:41 PM on July 11 and was still down on the 17th. (UTC is four hours ahead of Cuban time). Messaging apps WhatsApp and Telegram were also blocked, but not Facebook Messenger.

Cloudflare was also monitoring the situation in Cuba and, as shown below, there was a marked shift from mobile to desktop traffic. I read reports that mobile access had been cut off, and there are periods where 100% of traffic was from desktop users, but some mobile traffic moved during the protests.

Another interesting shift was in the percent of traffic written by humans to that from bots. Some of this shift may be a result of the blocking of human traffic or from increases in search engine activity. Since they report percents rather than absolute levels, it is hard to know.

In spite of blocking and suspension of Internet service in Cuba, there was a roughly three times increase in Cuban traffic during the protest time -- Cuba was in the news and Cubans were doing their best to communicate.

When blocked, many Cubans accessed the global Internet using the Psiphon VPN service. As shown below, the number of VPN users grew steadily during the protest period and reached a peak of 1.389 million daily unique users on July 15. I bet many new people learned about VPNs and learned to use Psiphon during the days of protest.