Chinese social media platforms add hidden digital watermarks to track screenshot sharing
Originally published on Global Voices
Image created from a screenshot of Douban.com
At least two Chinese social media platforms have added a script to their sites and applications to generate a hidden digital watermark (盲水印) to track the original user who shared a screenshot.
A digital watermark is a digital signal encoded and embedded in audio, video or image data. It is usually used to identify copyright. However, the technology seems to have been tweaked for censorship purposes in China.
Due to excessive censorship in China’s online public domain, many users have shifted to less public social media platforms to exchange their views. When users see important information or interesting discussions, they often take screenshots and share the images on public platforms, including Twitter. The practice makes it difficult for the censorship authorities to trace who originally leaked the information.
The watermark encoding script in a website or an application can insert a hidden digital code into a screenshot image. The code contains information about the source of the image, timestamp and user who took the screenshot. The hidden watermark can be revealed by adjusting the image's colouring.
The first social media platform that added the blind watermark script was Douban, a 200-million-user networking platform for sharing content related to film, books, music and news. Many have used the platform to create study groups. Feminist and LGBTQ+ groups are very active on Douban, and it has been a popular site for activism and subcultures, such as the lying flat movement, making it a target for crackdowns.
In December 2021, Beijing ordered app stores to remove more than 100 applications, including Douban, citing data privacy and security concerns. A month before members of the special task force from China’s Cyberspace Administration visited the Douban office to “rectify serious online chaos” in March 2022, the platform added the hidden watermark feature to its online group chat application.
According to Douban's official explanation, the auto-generated watermark includes information on the ID of the user who performs the screenshot function, the ID of the user who owns or publishes the content and the timestamp of the screenshot.
An initial tech analysis shows that the user footprints in the watermark cannot be erased by commonly used post-screenshot editing such as image rotation, cropping, resizing, pixellating and contrast adjustment.
While some have pointed out that the watermark feature would deter social media users from bridging information from one platform to another, because the Douban group chats are considered private, it has not triggered much controversy.
However, internet users were alarmed when Zhihu, a question and answer site, started testing the blind watermark feature in early September 2022. Twitter user @inroading revealed the hidden code by adjusting the colour of his screenshot:
继豆瓣后，知乎也全站（电脑端和APP）加盲水印了，截图分享是一件风险很高的事情 pic.twitter.com/qv40fSXt7e
— 阿良 (@inroading) September 4, 2022
After Douban, Zhihu (website and mobile app) also added the blind watermark feature. Now screenshot sharing has become a risky endeavour.
Zhihu seems to have suspended the watermark script after users raised their concerns. However, many are worried that the Chinese censor will eventually enforce the features on major Chinese social media platforms.
Meanwhile, many are sharing ways to circumvent the watermark feature. Here is one simple suggestion:
It is easy to hide traces. Step 1: Take a screenshot. Step 2: Open the screenshot in the system album, screenshot again. https://t.co/ochRV8EVpi
— Huazong (@realhuazong) September 4, 2022
Originally published in Global Voices.
