Facebook Reveals Coordinated Inauthentic Activity Takedowns in China, Philippines

The first network used a form of artificial intelligence known as generative adversarial networks

An example of content that was removed in the Philippines Facebook
Headshot of David Cohen

Facebook removed two separate networks for violating its policy against coordinated inauthentic behavior, with one originating in China and the other in the Philippines.

Head of security policy Nathaniel Gleicher wrote in a Newsroom post, “In each case, the people behind this activity coordinated with one another and used fake accounts as a central part of their operations to mislead people about who they are and what they are doing, and that was the basis for our action. When we investigate and remove these operations, we focus on behavior rather than content, no matter who’s behind them, what they post or whether they’re foreign or domestic.”

The first network was made up of 155 accounts, 11 pages, nine groups and six Instagram accounts that originated in China and focused primarily on the Philippines and Southeast Asia, and also on the U.S.

Posts in Chinese, English and Filipino in Southeast Asia focused on global news and current events including: Beijing’s interests in the South China Sea; Hong Kong; content supportive of Philippines President Rodrigo Duterte and Sarah Duterte’s potential run in the 2022 presidential election; criticism of Rappler, an independent news organization in the Philippines; issues relevant to overseas Filipino workers; and praise and some criticism of China.

In the U.S., content was posted both in support of and against presidential candidates Joe Biden and President Donald Trump, as well as former candidate Pete Buttigieg.

Gleicher said the network took operational security steps to conceal its identity and location, including the use of virtual private networks.

Graphika, a social analytics company with a focus on misinformation, worked with Facebook on analyzing this network, and it added more details in a report of its own, writing, “One way that the operation tried to get around that risk (being detected due to use of other people’s profile pictures) was by using profile pictures that appeared to have been generated by the form of artificial intelligence known as generative adversarial networks. This form of AI is readily available online, and its use (or abuse) by covert operations has exploded in the last year.”

The company continued, “GAN-generated images are a reliable way of sidestepping the need to clothe a fake account in a stolen profile picture; as such, they defeat the traditional investigative technique of reverse-searching the image. However, current iterations of GAN-generated images are by no means foolproof. The technology struggles with peripheral features, especially ears, glasses and hair, and it struggles still more with photo backgrounds since there is far more variation in the range of possible backgrounds a person can have than there is variation in the layout of the human face. At present, such indicators as these can usually reveal GAN-generated profiles to the naked eye. According to Graphika’s count, the Chinese operation used one-dozen such pictures among its accounts. Some appeared to have been cropped or to have partisan stickers superimposed on them, perhaps as a way to mask their more obvious deficiencies.”

Gleicher said about 133,000 accounts followed one or more of the pages, while around 61,000 people joined at least one of the groups and roughly 150 accounts followed one or more of the Instagram accounts. Approximately $60 was spent on advertising, paid for in Chinese yuan.

Examples of content from these pages and accounts follow:

The second network was made up of 57 accounts, 31 pages and 20 Instagram accounts that originated in the Philippines and focused on domestic issues.


david.cohen@adweek.com David Cohen is editor of Adweek's Social Pro Daily.
{"taxonomy":"default","sortby":"default","label":"","shouldShow":"on"}