Tuesday, July 18, 2017

China’s Censors Can Now Erase Images Mid-Transmission

China’s Censors Can Now Erase Images Mid-Transmission

Internet police step up their ability to filter photos

displays this...

BEIJING—China’s already formidable internet censors have demonstrated a new strength—the ability to delete images in one-on-one chats as they are being transmitted, making them disappear before receivers see them.

The ability is part of a broader technology push by Beijing’s censors to step up surveillance and get ahead of activists and others communicating online in China.

Displays of this new image-filtering capability kicked into high gear last week as Chinese dissident Liu Xiaobo lay dying from liver cancer and politically minded Chinese tried to pay tribute to him, according to activists and a new research report.

Wu Yangwei, a friend of the long-jailed Nobel Peace Prize laureate, said he used popular messaging app WeChat to send friends a photo of a haggard Mr. Liu embracing his wife. Mr. Wu believed the transmissions were successful, but he said his friends never saw them. “Sometimes you can get around censors by rotating the photo,” said Mr. Wu, a writer better known by his pen name, Ye Du. “But that doesn’t always work.”

There were disruptions on Tuesday to another popular messaging app, Facebook Inc.’s WhatsApp, with many China-based users saying they were unable to send photos and videos without the use of software that circumvents Chinese internet controls. Text messages appeared to be largely unaffected.

WhatsApp, which employs encryption that allows users to have secure conversations, is one of the few foreign messaging apps that has gone unblocked in China. Supporters of Mr. Liu had been using it to exchange information and images of the Nobel laureate in recent days.

Facebook didn’t immediately respond to a request for comment.

Chinese internet censorship first concentrated on the development of word-screening software to root out politically objectionable content. As a result, internet users over the past couple of years turned to sending photos to evade cyber police. In response, censors upped their game by demonstrating the ability to purge images from group chats and public posts.

In a new report, researchers from the University of Toronto’s Citizen Lab said they observed that WeChat expanded its image censorship to one-to-one chats for the first time, in the wake of Mr. Liu’s death on Thursday.

Citizen Lab said it is investigating how WeChat is able to filter the images. Since the images are blocked mid-transit, the speed is too fast for human intervention. The rapid blocking suggests an algorithm is at work, Citizen Lab researcher Lotus Ruan said.

Though activists said they noticed image censoring over the past year, Ms. Ruan said Citizen Lab didn’t detect this kind of targeted, person-to-person image blocking when it was investigating Chinese censorship in the spring.

Tencent Holdings Ltd., the Chinese internet company that operates WeChat, didn’t respond to requests for comment. Companies are required by law to maintain strict censorship of their platforms, a front-line defense that is augmented by police forces dedicated to internet monitoring.

Because WhatsApp is a foreign service, Chinese authorities can’t order it to implement fine-toothed censorship. Instead, to stop the transmission of certain images on WhatsApp, they have to disrupt the sending of images for all users in China.

The use of enhanced image filtering comes as Chinese authorities move to step up surveillance by using new data-driven technologies. Security cameras with facial recognition software are being deployed in Chinese cities to catch jaywalkers and track criminal suspects. Local governments are rolling out “social credit” systems that catalog the digital lives of its citizens, ranging from their internet history to bill payments.

A high-stakes leadership shuffle in the fall that is likely to see Chinese President Xi Jinping further consolidate his power could also help explain the evolution in censorship. Internet controls tend to tighten significantly in the run-up to major political events in China.

These new capabilities are closing a gap in censorship that Chinese activists and ordinary internet users have counted on—that the sheer mass of messages was too much for censors to handle.

“If you hire a million network police, it still wouldn’t be enough to filter 1.4 billion people’s messages,” said Bao Pu, a Hong Kong-based publisher of political books that are banned on the mainland. “But if you have a machine doing it, it can instantly block everything. It doesn’t matter if it’s a billion messages or 10 billion.”

Citizen Lab researchers said an increase in image censoring was noticeable as part of a broader clampdown on messages about Mr. Liu. The researchers said they documented 19 images blocked in one-to-one chats, including a cartoon of an empty chair, in addition to images blocked in group chats. Mr. Liu was famously represented by an empty chair at the 2010 ceremony where he was awarded his Nobel Peace Prize.

In tests conducted by The Wall Street Journal, some images of Mr. Liu were blocked in private WeChat messages, including a widely circulated one of him and his wife and another one overlaid with information about a vigil in Hong Kong. Some other photos transmitted successfully.

Activists said that they have noticed more frequent image blocks on WeChat over the past year and that there are signs the censorship is automated: one image will be blocked while a similar one in a different color scheme will go through.

Citizen Lab researcher Jeffrey Knockel said slight changes to an image or its metadata allow it to slip through the filter, while other modified pictures get blocked. That suggests WeChat is filtering based on certain data, or “hash,” of the image, Mr. Knockel said.