After accusations, Twitter will pay hackers to find biases in the automatic image crops

Twitter is hold a competition in hopes that hackers and researchers can identify biases in the image cropping algorithm — and it’s going to give out cash prizes to winning teams (through Engadget). Twitter hopes that by giving teams access to the code and model for cropping images, they can find ways in which the algorithm can be harmful (such as cropping in a way that stereotypes or erases the subject of the image).

Those who fight will submit a description of their findings, and a dataset that can be run through the algorithm to demonstrate the problem. Twitter will then assign points based on what type of damage is found, how much it could potentially affect people, and more.

The winning team will receive $3,500, and individual prizes of $1,000 will be awarded for the most innovative and most generalizable findings. That amount has caused some controversy on Twitter, with a few users saying it should have an extra zero. For context, Twitter’s normal bug bounty program would pay you $2,940 if you find a bug that allows you to perform actions for someone else (such as retweeting a tweet or image) using cross-site scripting. Finding an OAuth problem that lets you take over someone’s Twitter account would set you back $7,700.

Twitter has previously conducted its own investigation into its image cropping algorithm — in May, it published a paper examining how the algorithm was biased, following accusations that crop previews were racist. Twitter has since been largely deprecated with algorithmic cropping examples, but it is still used on desktops and a good cropping algorithm is useful for a company like Twitter.

Opening a contest allows Twitter to get feedback from a much wider range of perspectives. For example the Twitter team held a space to discuss the contest in which a team member said he was getting questions about caste-based biases in the algorithm, something that software developers in California may not notice.

It’s not just unconscious algorithmic bias that Twitter is looking for, either. The rubric has point values ​​for both intentional and accidental damage. Twitter defines accidental damage as crops that can result from a “well-intentioned” user posting a normal image to the platform, while intentional damage is problematic cropping behavior that can be exploited by someone posting maliciously designed images.

Twitter says in its announcement blog that the competition is separate from his bug bounty program – if you submit a report on algorithmic bias to Twitter out of competition, the company says that your report will be closed and marked as not applicable. If you are interested in participating, you can go to the HackerOne page of the contest to see the rules, criteria, and more. Entries are open until Aug. 6 at 11:59 p.m. PT and the challenge winners will be announced at the Def Con AI Village on Aug. 9.