McDonald’s hiring platform exposes 64M job applications

McHire, McDonald’s hiring chatbot platform, was protected by a default “123456” password, allowing researchers to access the admin side and every chat interaction that has ever applied for a job at McDonald’s.

There’s a reason why the infosec community relentlessly preaches about password security. Case in point: McHire. Security researchers Ian Carroll and Sam Curry recently published a blog explaining how they accessed the fast food chains’ hiring platforms chatbot with inexcusable ease.

“During a cursory security review of a few hours, we identified two serious issues: the McHire administration interface for restaurant owners accepted the default credentials 123456:123456, and an insecure direct object reference (IDOR) on an internal API allowed us to access any contacts and chats we wanted,” the researchers explained.

The pair focused their attention on McDonald’s AI hiring chatbot “Olivia,” after reading Reddit comments pointing to the chatbot giving users nonsensical answers. However, after trying to apply for a job at a local McDonald’s via the platform, Caroll and Curry noticed McHire’s admin website for restaurant owners.

Unexpectedly, after entering “123456” as the username and “123456” as the password, researchers were able to access the applications test area set up by McHire’s Ai chatbot creators, Paradox.ai.

“It turned out we had become the administrator of a test restaurant inside the McHire system. We could see all of the employees of the restaurant were simply employees of Paradox.ai, the company behind McHire,” researchers said.

While inside the test restaurant, Caroll and Curry looked into the chatbots’ application programming interface (API), noticing that there was a numbered ID for each applicant. Behind the ID numbers revealed previous applicant data with personally identifiable information (PII). Based on the number of IDs, the duo deduced the platform had over 64 million applications.

In theory they could have accessed all applicants’ data, such as:

• Name, email address, phone number, address
• Candidacy state and every state change/form input the candidate had submitted (shifts they could work, etc)
• Auth token to log into the consumer UI as that user, leaking their raw chat messages and presumably other information

The researchers have disclosed the issue with Paradox.ai, with the company solving the problem the next day.