Facebook fixed 61 high-severity flaws last year through its bug bounty program

Since 2011, the company has been paying monetary rewards to researchers who report flaws that could compromise the integrity or privacy of user data or could enable access to systems within its infrastructure.

While the minimum reward is US$500, there is no upper limit. The company decides how much to pay depending on a bug's severity and sophistication. The program doesn't cover only the facebook.com site and related services, but also other products that Facebook created or acquired, like Instagram, Parse, Onavo, Oculus, Moves and osquery.

In 2014, the company paid bug bounties totaling $1.3 million to 321 researchers from 65 countries, according to a newly published annual report. The average reward was $1,788 and the top three countries where valid bug reports originated were India, with 196 submissions; Egypt, with 81 and the U.S. with 61.

While Facebook did not reveal the largest bounty it paid last year for a single vulnerability, it pointed out that the top five earners collectively netted $256,750.

It's worth noting that, based on the statistics released by the company, finding a critical bug is not that easy. Facebook received 17,011 bug submissions in 2014 and those resulted in only 61 high-risk bugs being identified.

Unlike in previous years, Facebook didn't publish the total number of valid bugs that it identified over the course of last year as a result of its bug bounty program. In 2013 there were a total 14,763 submissions and 687 valid bugs, which would suggest that on average only 1 in 21 submissions leads to a new bug being discovered.

This also puts into perspective the resources needed for a company with a large website to run its own bug bounty program. There are bound to be many false positive, fake and duplicate submissions, which would require a large security team to sift through.

The program also helped Facebook identify some rather generic flaws that other developers out there might also have to deal with in their own sites and applications. The company gave three examples: a issue where backend code was receiving multiple values for the same parameter; an error where the attacker could register new S3 storage buckets on Amazon Web Services, which is used by many sites; and one that allowed legacy REST API calls to be made on behalf of users without proper authentication.