This article was originally published by Radio Free Asia and is reprinted with permission.
Researchers in the United Kingdom have discovered a coordinated network of “deep-fake” social media accounts pushing the Chinese Communist Party (CCP)’s views in the guise of ordinary account-holders.
The Centre for Information Resilience (CIR) said it had discovered a network of social media accounts that “distort international perceptions on significant issues, elevate China’s reputation amongst its supporters, and discredit claims critical of the Chinese government.”
In a report published on its website, the CIR said pro-China accounts were part of a “coordinated influence operation” on Twitter, Facebook, and YouTube that uses a mixture of artificial and repurposed accounts to post CCP propaganda.
“Our research shows evidence of a deliberate effort to distort international perceptions on significant issues – in this case, in favor of China,” CIR’s director of investigations Benjamin Strick said.
“There appears to be close overlaps in narratives shared by the network, to those shared by the social media accounts of [Chinese] state representatives and state-linked media,” Strick said.
Among the topics targeted are U.S. gun laws, COVID-19, overseas conflicts, and racial discrimination in an apparent bid to counter criticism of China, the report found.
CIR co-founder Ross Burley said the influence operation is similar to networks that have already been taken down by social media platforms in the past.
“It is likely that this operation is a continuation of those past efforts,” Burley said. “We urge the platforms mentioned in this report to investigate the network, formally offer attribution and take it down.”
The accounts identified in the report used machine-generated images as profile pictures, as well as images that appeared to be of real humans, and anime images as profile photos, Burley said.
“On Facebook and YouTube, many of the accounts also appeared to be repurposed,” Burley said. “There was evidence of previous authentic-appearing ownership of the accounts, indicating that at some point there was a change of ownership.”
‘Deep fake’ photos
At first glance, the profile photos in some of the fake Twitter accounts look ordinary enough, but the CIR was able to analyze them and discover that they had been generated by artificial intelligence.
The use of “deep fake” photos attached to pro-China accounts is a relatively new phenomenon, according to the report.
The pictures were made by stealing real-life profile photos and using a machine learning framework called StyleGAN.
The deep fake photos are generated by algorithms, so they are composed in a similar way.
For example, the eyes are always in the same position, and there may also be some very subtle defects, such as fuzzy hair, distorted limbs, or curled ears.
The report said it had identified hundreds of such accounts, which typically reuse the same hashtags, text, and photos, posting the same content on the same day.
Accounts suspended, removed
Twitter and Facebook said that most of the accounts reported to their platforms by CIR have been suspended and are being investigated.
In June 2020, Twitter removed 23,750 Chinese accounts linked to the CCP, which in turn fed into another 150,000 accounts that amplified China’s preferred narrative.
The tweets were predominantly in Chinese languages and spread geopolitical narratives favorable to the CCP, while spreading disinformation about Hong Kong, the IFJ said.
In August 2020, Google removed 2,500 fake YouTube accounts linked to China, with Facebook following suit in September 2020, taking down around 180 similar accounts, it said.
The Oxford Internet Institute (OII) has also published a report on Twitter accounts that supported the CCP line, although it was unable to prove any financial ties to the CCP.
Many of the accounts support the work of public figures engaged in pushing the CCP’s new brand of “wolf warrior” diplomacy on Twitter.
China’s ambassador to London, Liu Xiaoming, for example, had nearly 27,000 fake accounts following his account on Twitter, which had retweeted Liu’s tweets nearly 200,000 times before being deleted by the platform, the OII reported.
It said a batch of fake accounts pretending to be British citizens appeared on Twitter and started to promote CCP-friendly content, generating more than 16,000 retweets before being deleted by Twitter at the end of April 2021.