What is zero trust?
Zero trust is a security model in which no request is trusted because of where it comes from. A device on the office network, a familiar address, a session opened an hour ago: none of these is proof. Every access to a resource is checked against policy, each time it happens.
The term sits in the WAF glossary, and it is often misread as a product category. It is closer to a rule of design that the rest of the security stack is arranged around.
Quick facts
- Core idea
- No implicit trust from network location or asset ownership; access is decided per request or session
- Reference text
- NIST SP 800-207, Zero Trust Architecture, published August 2020
- NIST definition
- A set of paradigms that move defenses from static, network-based perimeters to users, assets and resources
- Tenets
- Seven, listed in section 2.1 of the NIST document
- Maturity guidance
- CISA Zero Trust Maturity Model, version 2.0, aligned to OMB M-22-09
- Not
- A single product, a one-time purchase or a replacement for application logic
- Related
- Endpoint security, BOLA, API authentication, JWT
Perimeter versus zero trust
The older picture is a castle and a moat. A firewall at the edge separates a trusted inside from a hostile outside. Anything that gets through the gate is assumed to be friendly, and traffic between internal machines is waved along. The design is easy to explain and cheap to run while everything worth protecting sits behind one wall.
It fails in a predictable way. One stolen password, one infected laptop or one forgotten server puts an attacker on the inside, where the checks are weakest. Applications moved to cloud services, staff work from home, and partners connect from their own networks, so the wall no longer encloses what it was built to enclose.
Zero trust drops the assumption that the inside is safe. A firewall can still be part of the design, and NIST says as much: many definitions of zero trust still describe themselves in relation to perimeters, for example through micro-segmentation. What changes is that standing inside the perimeter no longer earns trust on its own.
| Question | Perimeter model | Zero trust |
|---|---|---|
| What earns trust? | Being inside the network | Passing authentication and authorization for this access |
| When is it checked? | Mostly once, at the edge | Per session or per request, with the policy applied again |
| What is protected? | The network segment | The individual resource |
| After a break-in | The intruder moves freely inside | Each further step needs its own approval |
What NIST says
The document most often cited is NIST SP 800-207. It describes zero trust as a set of cybersecurity paradigms that move defenses away from static, network-based perimeters and toward users, assets and resources. A zero trust architecture applies those principles when planning enterprise infrastructure and workflows. The central statement is that no implicit trust is granted to assets or user accounts based on their physical or network location.
Section 2.1 lists seven tenets. NIST calls them the ideal goal and notes that not every tenet may be fully implemented in a given strategy. Paraphrased:
- All data sources and computing services are treated as resources.
- All communication is secured regardless of network location. Network location alone does not imply trust.
- Access to an individual resource is granted per session, with the least privilege needed, and access to one resource does not carry over to another.
- Access is determined by dynamic policy, which can draw on the identity of the client, the state of the requesting asset and behavioral or environmental attributes.
- The enterprise monitors and measures the integrity and security posture of the assets it owns or is associated with. No asset is inherently trusted.
- Authentication and authorization are dynamic and strictly enforced before access is allowed.
- The enterprise gathers as much information as possible about assets, network traffic and access requests, and uses it to improve policy.
The document also names the parts of its logical architecture. A policy engine decides whether a subject may access a resource. A policy administrator sets up or tears down the communication path. A policy enforcement point enables, monitors and eventually terminates the connection between the subject and the resource. NIST groups the engine and administrator as the policy decision point. The split matters here for one reason: the decision and the place where it is carried out are separate jobs, and a request-filtering layer can be at best one of the places where enforcement happens.
The CISA maturity model
NIST describes the principles. CISA's Zero Trust Maturity Model describes how far along an organization is. Version 2.0 is aligned to OMB memorandum M-22-09. The model has five pillars and three cross-cutting capabilities, and within each pillar it gives examples of traditional, initial, advanced and optimal zero trust architectures.
CISA's own description of zero trust is useful for its wording. It speaks of concepts and ideas designed to minimize uncertainty in enforcing accurate, least-privilege, per-request access decisions. It assumes the network may already be compromised, and it calls for a shift from a location-centric model to a more data-centric one. It also says that adoption may require a change in an organization's philosophy and culture around cybersecurity, which is a fair warning that the work is not mainly technical.
Read the model as a way to place your own setup on a scale, not as a certification. It is written for federal agencies, and a small site does not need to copy its structure to take its main point.
What it is not
It is not a product. Vendors use the phrase on many things, but NIST and CISA both describe principles and a direction of travel. A tool can support a tenet. It cannot be the model.
It is not a single purchase or a project with an end date. The tenets include monitoring and continual re-evaluation, which are ongoing work.
It is not the abolition of firewalls, networks or segmentation. And it does not mean distrusting people. The distrust is aimed at the assumption that a location or an earlier login settles the question, not at the user.
Zero trust for web apps and APIs
A web application is a small version of the same problem. Requests arrive over the public internet, and where they come from says almost nothing about whether they are allowed. Applying the tenets at that scale comes down to a few habits.
Check authentication on every request, not only at login. A session cookie or token is a claim that the earlier login happened, and a stolen one lets an attacker replay it, which is how session hijacking works. Short lifetimes and re-checks on sensitive actions follow tenets 3 and 6. For tokens in particular, see JWT and API authentication.
Check authorization on every route and every object. Tenet 3 says access to one resource does not imply access to another, and in an API that is the difference between a login and a permission. When an endpoint takes an object ID and returns the record without asking whether this caller owns it, the result is broken object level authorization. Each route also needs its own check, which is the subject of endpoint security.
Give each caller the least privilege that the task needs, and log what happened. Tenet 7 asks for data about access requests so policy can improve. For an application, that means knowing which routes get hit, by whom, and with what result. The wider topic is covered under API security.
Where a WAF helps and where it does not
A web application firewall sits in the request path, often as a reverse proxy, so it sees every HTTP request before the application does. That position suits the tenets about checking each access and collecting data. It can judge the content of a request, apply limits and keep a record of what it flagged. In the language of NIST's model, it can be one of the points where a policy is applied.
It cannot do the application's part. A WAF does not know who your users are, which records belong to whom, or whether a valid token should be allowed to open a particular object. A request for someone else's invoice can look exactly like a request for your own. That decision needs the application's data model, which is why authentication and authorization stay in the application and a filter does not replace them.
Bridge WAF has three modes: off, monitor and block. Start in monitor, read the activity log and see what real traffic does before blocking anything. That order suits a zero trust project too, because the first job is to learn what is reachable and who is using it. Nothing here means Bridge WAF delivers zero trust. It is one layer of filtering in front of an application that still has to check its own permissions.
Questions
What is zero trust in simple terms?
Zero trust means a request is not trusted because of where it comes from. Being inside the company network, or holding an address that used to be safe, grants nothing. Each access to a resource is checked on its own: who is asking, from what, and whether policy allows it right now.
Is zero trust a product I can buy?
No. NIST describes it as a set of principles for planning infrastructure and workflows, and CISA calls it a collection of concepts and ideas. Products can help apply those principles, but no single purchase makes a system zero trust.
Does a WAF give you zero trust?
No. A WAF sees every HTTP request that reaches it and can judge each one, so it can be one point where policy is applied. It does not know who your users are or which records they may open. Authentication and authorization still belong in the application.
How does zero trust apply to an API?
Check authentication on every request and check authorization on every route and every object, with the least privilege that the task needs. A valid login or token proves who is calling, not what that caller may read or change.
Where should I start with zero trust on a web application?
Start by finding out which routes are reachable and which of them check permissions. Then close the gaps in the application. A filter placed in monitor mode first can show what actually arrives before anything is blocked.