It goes without saying that companies migrating IT systems and operations to the cloud face a growing number of challenges related to security. From securing cloud-native applications in a continuous release environment to managing digital identity and remote access to sensitive data, virtually every aspect of day-to-day IT and cybersecurity operations is changing rapidly. 

With Gartner forecasting $124 billion in worldwide cybersecurity spending this year, it’s clear that organizations are racing to find a model that works. There’s just one problem: far too many are trying to shoehorn legacy security models into a cloud-native world – and it’s not working. In fact, some of the best security teams in the world today even struggle to keep their systems secure, including Facebook, Google, Microsoft and a growing list of others. 

What’s going on here? It’s quite simple, actually: Just as the move to cloud forces organizations to undergo fundamental shifts in their business and IT operations, it also demands they make equally substantial changes to their core assumptions about cybersecurity. 

In response, many IT and security experts are quick to proclaim, “the perimeter is dead.” While it’s certainly true that the old dichotomies of inside/outside or attackers/trusted parties have become less clear, the notion that perimeters are altogether gone is a dangerous misconception. In today’s cloud-first world, the perimeter isn’t gone – it’s ephemeral. Instead of being fixed and centrally organized, perimeters are just like every other aspect of the cloud now — they are distributed and ever-shifting as conditions change. 

What’s needed is a cloud-native security model. The good news is that it already exists, and it’s been around for nearly a decade: zero-trust. 

Zero trust is a radical change from pre-cloud security models. Instead of assuming it’s possible to create a central perimeter that protects everything inside (the “hard shell, soft center” approach), zero trust is all about: (1) reducing an organization’s security risk by minimizing the probability of an attack via proper identification and authentication, and (2) minimizing the impact of successful attacks by microsegmenting authorization. That is, if someone steals a user’s credentials, they can no longer slip through the organization’s defenses undetected. As the user moves through a system, they must continuously verify that they are who they say they are, with the level of verification required scaling alongside the sensitivity of the access and the task they are requesting or performing. 

Where do you start? While every organization is different, there are three core principles that guide all zero trust implementations: 

There is No Distinction Between “Insiders” and “Outsiders”

While most organizations today recognize that the line between “insiders” and “outsiders” is blurring, most still follow the same model and view an employee on an “authenticated” device as trusted. But BYOD, phishing and rogue insiders demolish that distinction.Everyone should be considered hostile. No implicit trust should be attributed to the location of the user, the network or the device they are using. 

For example, Mobile Device Management (MDM) tools give the IT team control over the use of a personal device in the network like a tablet, but doesn’t provide true security, only the illusion of it. If the user clicks on something that the MDM agent on the device can’t see, the network can be easily compromised. For instance, employees in the HR department receive and open emails with attachments containing resumes all the time. However, MDM doesn’t protect against attachment-borne malware that installs a trojan agent on the employee’s device. If the security control is to treat every device under management as trusted, then this trojan agent can simply hide in plain sight.  

Discrete One-time Super Authentication Must End

Most organizations have implemented 2FA, which is good, but the problem is its use is too far reaching. Once a user is verified with 2FA they can access all sorts of systems because the model considers their identity as confirmed going forward. However, identification and authentication are contextual, and the results should be valid only for the duration of the specific transaction for which they are requested. One successful spear phishing campaign can turn a user’s device from safe to risky if malware is downloaded on it.  Very few organizations do transaction-level verification. Banks do it for certain types of high-level transactions, but outside of that, it’s rare. Two-factor authentication shouldn’t grant you elevated privileges indefinitely; it’s a discrete event. 

We understand this in other contexts. For instance, when flying. You don’t just buy a ticket and then walk right onto the plane. You have to check in at the airline desk, go through security with your boarding pass and ID, and then show your boarding pass again at the gate. Each new level of verification allows deeper access but not carte blanche access. A boarding pass in conjunction with a valid ID gets you past security, but not into sensitive areas of the airport. Similarly, a boarding pass gets you onto a plane, but does not allow you to access the flight deck. Another example is the medical field. Certain prescriptions, like narcotics, require patients to provide ID and take extra steps to ensure against fraud. 

Use Dynamic Policies With Broad Inputs

Too often, organizations treat policies as static and rely on too few inputs, partially out of habit and partially due to historical limits on compute power and inability to ingest and process a multitude of signals. But to be effective policies need to be dynamic and to be evaluated with the largest possible set of inputs. This means that trust in the identification and authenticity of the user, and the risk associated with the task they are attempting to perform, are constantly evaluated.

For example, you can have static policies that say if a user is authenticated, then let them in. Or, if a user is authenticated and has done biometric, let them in. And, if the user is authenticated but it’s after 5 pm, ask for biometric then let them in. Requests from managed devices provide more inputs to use to establish the level of trust. But static policies break because they cannot process new information or interpret new context, so they fall back to a default rule which might or might not be the right approach.  Dynamic policies allow systems to act on incomplete information and provide an answer that is a trust level on a continuum, rather than binary yes/no trust level. 

Such security policies aren’t reliable; their lack of sophistication enables a false sense of security and/or interferes with productivity needlessly. As humans, we make dynamic decisions and judgment calls all the time — walking down the street and assessing the riskiness of a neighborhood at certain times of day, or deciding what to wear by checking the weather forecast and looking at the sky. Security policies need to allow for similar degrees of input and to reflect the reality of changes in users, devices and contexts. 

In a hyperconnected world, there is no such thing as absolute security and zero risk. Now more than ever, security is about continuously evaluating trust and making decisions to minimize risk. Zero trust principles offer clear and actionable guidance. They ensure protection of data no matter what users or devices are attempting to access them by operating under the assumption that all devices and systems might be compromised.  

Baber Amin, CTO West, Ping Identity