'Heisenberg Cloud' Spots Exposed Database Services, Misconfigurations

A Rapid7 research project to better understand the cloud threat landscape has unearthed some interesting findings.

Prime among them are the extent to which services like databases are deployed directly to the public Internet space in the cloud and the amount of misconfiguration in legitimate cloud and Web application services.

The Rapid7 project, dubbed Heisenberg Cloud, combines data from a honeypot framework developed by Rapid7 called Heisenberg with scan data collected under a separate Internet-wide project Rapid7 heads up, Project Sonar.

The effort is to build a repository of honeypot data to help security researchers measure the extent and types of attacks and researcher activity in cloud provider environments.

Much of the data gathered under the project in the initial phase is from honeypot agents deployed in cloud environments from six major cloud services—Amazon Web Services, Google Cloud, Rackspace Cloud Hosting, Microsoft Azure, IBM SoftLayer, and DigitalOcean.

In the column of expected findings were things like the sheer number and variety of attacks directed at the honeypots. Examples included attempts to exploit PHP services, desktops, point-of-sale system databases, and a recently disclosed National Security Agency backdoor in a Juniper firewall. Also unsurprising was the discovery that between 50- and 80% of customer nodes expose Web services, according to Rapid7.

But the volume of database services exposed directly on the Web was surprising, says Bob Rudis, Chief Security Data Scientist at Rapid7. For instance, 22%, or more than 1 in 5, SoftLayer customer nodes expose MySQL and SQL Server database servers directly on the Net.

The Heisenberg Cloud project found a number of misconfigured cloud services among companies using dynamic cloud infrastructure components, Rudis says. For example, there were legitimate cloud services attempting to talk to infrastructure that no longer exists due to stale configurations either hardcoded in them or in their DNS entries.

“Also, while we knew there was a large amount of anonymous proxy traffic, we didn’t really expect to see the volume of both illicit content scrapers or legitimate companies using anonymous proxies to illegally harvest content from sites like airlines,” Rudis says.

The Heisenberg Cloud project is still only in its initial stages, but already there are some key takeaways for enterprises using cloud services, the Rapid7 report said..

For instance, the study showed a disproportionate amount of attacks, anonymous proxy probes, and misconfigured traffic coming at TCP ports 1080, 3128, 8000, 8080, 8083, 8118, or 8888.

Organizations using cloud service providers should avoid using these ports, Rudis says.

The ports identified in the report, especially those in the 8000, tend to be default configurations of various popular Web and API services. “Many organizations deploy these directly to the Internet with no filtering and with these default configurations,” he says.

Given the diversity of the volume and velocity of the communications on these ports, there is a real danger to organizations from using them, even if it is just from a stability standpoint, he says.

Where possible, organizations should consider using private networking and Virtual Private Clouds to keep data and infrastructure away from direct Internet access. Another good idea is to log everything and to analyze everything you log, Rapid7 notes.

Rapid7 plans to deploy specialized honeyports to fingerprint actual attacker behavior, Rudis says. The company will also work with its enterprise customers to deploy the honeypots inside their networks to see how it differs from the cloud threat environment.

Related stories: