Shay Nahari, who heads the Red Team at CyberArk, said in a blog post that domain fronting had been used before in the wild by experienced attackers as an evasion technique.
Nahari told iTWire that the company had identified tens of thousands of domains served by Akamai's content delivery network that could be used for domain fronting. These included Fortune 100 domains and could be used in targeted attacks.
According to a research paper, domain fronting uses different domain names at different layers of communication. In an HTTPS request, the destination domain appears in the DNS query, the TLS server name indication extension and the HTTP host header.
The frontend server would direct the request to the domain specified by the host header and no traffic would go to the domain specified by the other two parameters.
Nahari said that for the CyberArk method to be used, one needed a content delivery network that hosted a two-way persistent read-write system or application.
"For example, in our testing, our malware utilised the API of a commercial to-do list application hosted by the content delivery network to exchange instructions and data with us," he said.
Also needed was malware that had been specially written to use this command-and-control channel; it needed to be implanted on users' machines.
Nahari said multiple third parties' API sharing the same edge network were first identified within the content delivery network, and multiple persistent lists of commands were created using the shared API interface. These lists acted as the malware command queue. Different lists were utilised by the malware and the server to issue commands and receive output.
Custom malware was then created to use token data and the content delivery network customer's API to read and write requests to the command-and-control system.
From the perspective of the defender, the client machine would be sending a request to a reputed domain and the traffic would be encrypted and signed by this domain, he said.
One way of mitigating this method of attack was for organisations to monitor and control traffic leaving their networks, Nahari said.
"One theoretical solution is for the content delivery network to associate virtual IP addresses with each domain, which can be assigned to a specific SSL certificate and therefore protect against the cross-tenant approach we’ve demonstrated," he said.
"There are, of course, practical limitations to this approach, perhaps most importantly the shortage of public IPv4 addresses."
Nahari said while Akamai's was not the only content delivery network that was susceptible to this technique, the fact that it handled so much Internet traffic made it a juicy target.