Project #5 Anti-Surveillance Tool
Design Process
The idea I choosed to develop in this project is text-image conversion as a means to evade text censorship. Using techniques found in CAPTCHA image generation, text and/or image can be disguised and distorted in a certain fashion that greatly reduce machine-recognitability but not to human readers.
Significance & Context
Chinese social media automatically censors sensitive keywords, especially those discussing the government, politcs or thorny social issues. Massive posts are deleted and accounts banned during events like the COVID-19 outbreak in Wuhan. No news or discussion allowed around the situation during which the scenario was first emerging, and this problematic approach of censorship led to devastating results. This tool that I developed can convert paragraphs of text into CAPTCHA-like image that is difficult for machines to censor.
Even though emerging technologies such as machine learning based Optical Character Recognition have achieved unprecedented accuracy, this project is still meaningful and effective in the way that it exponentially increased the strain on processing power were the authority decide to crack and censor these imaged texts.
Reflection
The images generated by the current version of this tool is no where near perfect. Texts are overflowing the canvas, fonts are too limited and some are difficult to recognize even by human eyes. The user also have to write the paragraph and save the image then send through their chosen medium. This is, to me at least, a chore to do if I want to use it regularly. Maybe a plugin for web-based social media could be a better route to go for more usability.