Facebook has shut down more than 180 fake accounts, groups, pages, and Instagram accounts that it determined to be run in China, which posted content on the U.S. presidential election and spread Beijing’s talking points on a range of topics, from the South China Sea to Hong Kong protests.
In total, 155 Facebook accounts, 11 pages, nine groups, and six Instagram accounts were shut down. The Instagram app is owned by Facebook.
Nathaniel Gleicher, Facebook’s head of security policy and author of the blog post, explained that while people behind these accounts tried to conceal their identity and location, including by using virtual private networks (VPNs), the company was able to trace the account operators to southern China’s Fujian Province.
Content
According to Facebook, this Chinese-created network focused most of its posts on Southeast Asia, posting news about global events such as Chinese and U.S. naval activities in the South China Sea, protests in Hong Kong, and support for current Filipino President Rodrigo Duterte and the possible 2022 presidential bid of his daughter, Sara Duerte, who is currently the mayor of Davao city in the Philippines. People behind these accounts posted in Chinese, Filipino, and English.On the upcoming U.S. election, these accounts posted content both “in support of and against presidential candidates Pete Buttigieg, Joe Biden, and Donald Trump,” Gleicher stated.
In terms of the campaign’s appeal, about 133,000 Facebook accounts followed one or more of these fake Facebook pages, and around 61,000 people joined one or more of these groups. However, the network “gained almost no following” in the United States, according to Gleicher.
Facebook also found a second inauthentic network, originating from the Philippines. The firm took down 57 accounts, 31 pages, and 20 Instagram accounts in connection with it.
Investigation
U.S. network analysis firm Graphika released its own investigative report on the recently deleted Chinese-linked accounts, revealing more information about their tactics. Graphika is commissioned by the social media company to conduct such analyses.It found that the operation behind these accounts began in late 2016 and it repeatedly went back to the “theme” of maritime security, in particular, “the achievements of the Chinese navy.”
For example, one Facebook account named “Modern Chinese Warship” posted on March 21, 2018, about the recent return of the Chinese navy after completing a mission in the Gulf of Aden.
When the operation began in 2016, a cluster of accounts focused on Taiwan. One Facebook page named “Things About the Taiwan Strait” first began posting “comments about the prowess of mainland China and the benefits that Taiwan would reap from reunification,” according to the report.
According to Graphika, the Facebook group changed its name to “For Pete Buttigieg” on July 4, four months after he dropped out of the presidential race.
Graphika also found that the network had begun creating accounts and pages “with apparently American names and personas” by April 2019. For example, one account, under the name Kate Selina, posed as a conservative American, with posts criticizing Medicare for All and gun-control policies.
Another account, named Brian E. Gerald, posed as a liberal American, posting memes mocking President Donald Trump.
The network also included a Facebook group named “Trump KAG 2020” that supported Trump, and a group called “Biden Harris 2020” that supported the Democratic presidential running mates.
Only the Biden–Harris page had a significant following before it was taken down by Facebook. According to Graphika, it had about 1,400 members.
The firm gave two hypotheses as to why the operation posted such content.
The firm also surmised that the network’s intention could be “to use election-related messaging to infiltrate online communities on both the right and the left, especially those communities that were interested in the U.S. Navy and maritime issues more generally.”
Masking Fake Accounts
Graphika also found evidence to suggest that the network might have “purchased or otherwise acquired” some of its accounts from real users, after observing that some of the profile names differed from the name that appeared in its URLs. Facebook generates a unique URL for each user’s profile, which usually contains the user’s name.Some accounts also posted photos generated by a form of artificial intelligence (AI) known as generative adversarial networks (GANs) to avoid being detected as fake accounts.
Doing a reverse search on photos can expose accounts that use borrowed pictures, which confirms they are fake.
“This form of AI is readily available online, and its use (or abuse) by covert operations has exploded in the last year,” the Graphika report stated.
However, these images aren’t “foolproof,” as there are subtle anomalies that can be detected with the human eye, such as distortions in the ears and hair.
For example, in the photo of an account named Polo Kalia Sota, the photo background is distorted, and the glasses and ears appear asymmetrical.
Graphika identified a dozen of these GAN-generated images among the network’s accounts.