Broker injection attack is a type of vulnerability that exploits misconfigured brokers, potentially allowing an attacker to read, write and inject information from/into their flow.

There are many scenarios in which a broker is used to transport the information between tasks.

One of the most typical use cases is send e-mails in background. In this scenario we'll have two actors:

• An information producer (a website, for example).
• A worker or background process who actually sends the e-mail.

The producer needs an asynchronous and non-blocking way to send the email information to the worker.

This system is usually a broker. It takes the information from the web front-end and passes it to the worker, generating a new task in the worker. So, the worker has all the information to send the e-mail.

Taking the above scenario as an example, if we could access the broker, we would be able to make the worker generate new tasks with arbitrary data, unleashing a broker injection.

With this in mind, we could make the following attacks:

• Listing remote tasks.
• Reading a remote task's contents.
• Injection of tasks into remote processes.
• Removing remote outstanding tasks.

The broker injection attack is not new, but it didn't have a name. This name was coined by Daniel García (cr0hn) at the RootedCON 2016 conference in Spain.

• Redis
• RabbitMQ
• ZeroMQ
• Message broker
• Celery (software)

• Official Redis security tips
• Enteletaor: The broker injection tool
• Broker injection in RootedCON 2016 (Spanish)^{[permanent dead link]}