Back 
Co-authored by James Robinson and Jeff Kessler
As rapidly as wide-area networking (WAN) and remote access strategies with associated technologies are changing, we’re always surprised by the amount of time some security professionals and auditors dedicate to the either/or debate between split tunnel and full tunnel connectivity.
History can partially explain how we got here. Long before COVID-19 reared its ugly head, corporate security teams were already grappling with how best to protect remote connectivity. First, they directed all employees working remotely to reach the corporate network via VPN, which meant all traffic was routed through the firewall/VPN concentrator in the enterprise data center. When these users were only, or primarily, accessing applications housed inside the corporate network, that made sense, although bandwidth limitations sometimes reduced application performance. A middle ground many of us found (we liked to call it splint-tunneling) was also to enable direct connections for “approved” cloud services while everything else was sent to the data center. However, this approach still had weaknesses including the abuse of these exposed cloud services and the lack of visibility by security teams. As cloud solutions became more and more prevalent, forcing traffic traveling from remote offices to cloud-based applications (and vice versa) to make a pitstop in the data center began to make less and less sense.
The easy solution was to reduce unnecessary backhauling to the data center by letting remote machines talk directly to the internet. Many security teams implemented “split tunneling,” by which traffic that needs to pass through the corporate network utilizes VPN connections, while traffic that is headed to the internet goes there directly without visiting the data center at all.
But easy is not security and split tunneling raises massive red flags for security teams. Internet browsers, software-as-a-service (SaaS) applications, and streaming technologies create new attack vectors for malware. Unless these solutions’ data streams flow through the corporate firewall, the company is relying solely on endpoint protection solutions for threat detection and mitigation. If a DNS or ICMP attack were to succeed in bypassing endpoint security, not only could it be used as a covert channel, but it could also be the entry point to the enterprise over the VPN connection. Historically security teams have not had good visibility into these kinds of attacks where the command and control traffic is asynchronous to the tunneled traffic that goes into the enterprise. Furthermore, these split-tunneled systems were mini-pivot points allowing the compromise of a system to be entry into data center secured systems and applications.
Split tunneling was alrea