Valtik Studios
Back to blog
Zero TrusthighUpdated 2026-04-17orig. 2026-02-0311 min

Zero Trust Architecture in 2026: Past the Buzzword, Into the Implementation

Zero Trust has been a marketing term since 2014 and a budget line item since 2020. Here is what actually ships in 2026, NIST SP 800-207 in practice, the vendor shootout (Zscaler, Cloudflare, Netskope, Cato, Palo Alto), and the implementation mistakes that turn Zero Trust into a VPN with a nicer logo.

TT
Tre Trebucchi·Founder, Valtik Studios. Penetration Tester

Founder of Valtik Studios. Pentester. Based in Connecticut, serving US mid-market.

# Zero Trust Architecture in 2026: past the buzzword, into the implementation

I'll say what most consultants won't. The thing your vendor is selling you is probably not Zero Trust. It's a VPN with an identity provider bolted on, wearing a new coat.

Forrester coined Zero Trust in 2014. COVID made it a budget line item in 2020. OMB M-22-09 made it a federal mandate in 2022. By 2026 every single security vendor has something with "Zero Trust" in the name. Maybe 30% of those products actually implement a Zero Trust architecture. The rest rebrand their existing product and hope nobody reads the NIST spec.

Here's what this post actually covers. How to tell the real implementations from the rebadges. What NIST SP 800-207 actually requires versus what vendors claim it requires. A rundown of the major vendors with honest strengths and weaknesses. And the mistakes we walk into on every Zero Trust audit engagement.

What Zero Trust is

NIST SP 800-207 defines Zero Trust as a security model where "no implicit trust" is granted based on network location. Access decisions are made per-request, per-resource, with continuous verification of identity, device posture, and context.

The BeyondCorp paper from Google (2014) is the original implementation pattern. Google's internal decision after the 2010 Aurora attack was: stop trusting the corporate network. Every request gets authenticated and authorized as if it came from the open internet. A laptop sitting in the office has no more trust than a laptop sitting in a hotel.

The three core principles:

  1. Verify explicitly. Every request is authenticated with identity AND device posture AND context (time, location, request pattern). Not "user logged in once this morning."
  2. Use least privilege access. Users and services get the minimum access needed for the specific task, time-bound where possible. No blanket VPN access to the flat corporate network.
  3. Assume breach. Design assuming the attacker is already inside. Segment aggressively. Log everything. Monitor for lateral movement continuously.

The architecture components

Policy Decision Point (PDP) and Policy Enforcement Point (PEP)

The PDP makes the access decision. It takes inputs (identity, device, context, resource) and outputs allow/deny plus conditions. The PEP sits in front of the resource and enforces what the PDP decided.

In practical deployments:

  • PDP = your IdP (Okta, Entra ID, Ping) + your policy engine (same IdP, or a separate service)
  • PEP = the proxy, the service mesh sidecar, the identity-aware proxy (IAP), the ZTNA gateway

Identity Provider

The single source of truth for user identity. Okta Workforce Identity Cloud, Microsoft Entra ID (formerly Azure AD), Ping Identity, OneLogin, JumpCloud, Google Workspace Identity. All apps, all access, all flows authenticate through the IdP.

Required capabilities:

  • MFA (phishing-resistant, preferably FIDO2/WebAuthn)
  • Risk-based adaptive access (location, device, behavior anomaly)
  • Session lifetime management (short sessions, forced reauth on sensitive actions)
  • SCIM provisioning to apps (automated deprovisioning when an employee is terminated)
  • SSO to all applications (no password-based direct logins)

Device Trust

A policy decision without device context is "user trust." Device trust requires:

  • Device inventory (MDM-enrolled or EDR-reporting)
  • Posture checks (OS version, patch level, disk encryption, EDR running, no jailbreak)
  • Device certificate issued during enrollment (proves device identity to the ZTNA gateway)
  • Continuous posture verification (not at login)

Common device trust stacks:

  • Jamf Pro / Kandji / Mosyle (macOS) + Intune (Windows) + Google Workspace Endpoint Management (ChromeOS/Android)
  • CrowdStrike / SentinelOne / Microsoft Defender for Endpoint. EDR that reports posture
  • Okta Device Trust, Jamf Connect, Google BeyondCorp Endpoint Verification
  • Device certificates via cloud PKI (Smallstep, EJBCA, AWS ACM Private CA, Microsoft ADCS)

Micro-segmentation

Replaces the flat corporate network with per-workload or per-app segments. Lateral movement requires explicit authorization.

Techniques:

  • Service mesh (Istio, Linkerd, Consul). MTLS between every service, per-service authorization policies
  • SDN with ACLs (Illumio, Guardicore/Akamai, VMware NSX)
  • Kubernetes NetworkPolicies (Cilium, Calico)
  • Cloud-native (AWS VPC Lattice, GCP Service Connect, Azure Application Gateway private endpoints)

Continuous Monitoring and Analytics

Zero Trust assumes breach. Every access event is logged. Behavioral analytics flag anomalies. Response playbooks can revoke sessions, step-up authentication, or quarantine devices automatically.

Typical stack:

  • SIEM (Splunk, Sentinel, Chronicle, Sumo Logic, Elastic)
  • UEBA (Exabeam, Varonis, Microsoft Defender for Identity)
  • XDR (Palo Alto Cortex, CrowdStrike Falcon Complete, Microsoft Defender XDR)
  • SOAR for response automation (Palo Alto XSOAR, Splunk SOAR, Tines)

The 2026 ZTNA vendor shootout

ZTNA (Zero Trust Network Access) is the market category for the replacement of traditional VPN. Here's what ships in 2026.

Zscaler Zero Trust Exchange

Pros:

  • Most mature platform, largest market share
  • Global edge footprint (150+ data centers)
  • Private Access (replaces VPN) + Internet Access (replaces SWG) + Digital Experience Monitoring
  • Strong integration with Microsoft Entra, Okta, CrowdStrike, SentinelOne
  • Dedicated federal cloud for FedRAMP High

Cons:

  • Expensive. Per-user pricing stacks fast.
  • Architecture is cloud-only. Latency matters for users far from a Zscaler POP.
  • Complex policy model. Requires dedicated ops staff to maintain at scale.

Best fit: enterprise (5K+ users), global distributed workforce, compliance-heavy industries.

Cloudflare Zero Trust (Cloudflare One)

Pros:

  • 300+ data centers globally, low latency everywhere
  • Bundle pricing. Access, Gateway, WARP, Tunnel, Browser Isolation in one SKU
  • Excellent for technical teams. Cloudflare Tunnel replaces VPN with outbound-only connections
  • Free tier for up to 50 users (Cloudflare Access)
  • Strong developer experience, Terraform-native

Cons:

  • Less mature on the "corporate" side (device management, HR integrations) compared to Zscaler/Netskope
  • UI can feel unpolished for non-technical admins
  • SSO enforcement on applications requires wrapping every app behind Access (no native SAML-replacement for non-HTTP apps without workaround)

Best fit: mid-market, technical organizations, cost-conscious enterprises.

Netskope One

Pros:

  • Strong on CASB (Cloud Access Security Broker). Deep inspection of SaaS usage
  • SSE (Security Service Edge) platform combining ZTNA + SWG + CASB + DLP
  • Good at DLP and data-centric policies
  • Solid integrations with major IdPs

Cons:

  • ZTNA side sometimes feels bolted on compared to Zscaler's focus
  • Pricing on the high end
  • Initial configuration complexity

Best fit: organizations heavy in SaaS usage, regulated industries with strict DLP requirements.

Palo Alto Prisma Access

Pros:

  • Integrates with Palo Alto firewalls. If you already run Palo Alto, smooth path
  • AI-powered threat detection across the network
  • Strong NGFW feature set in the cloud service
  • Solid FedRAMP posture

Cons:

  • Lock-in to Palo Alto ecosystem
  • Complex licensing model
  • Premium pricing

Best fit: existing Palo Alto customers, hybrid enterprises.

Cato Networks SASE

Pros:

  • Fully converged SASE platform (SD-WAN + ZTNA + SWG + CASB + FWaaS)
  • Single management plane for everything
  • Good performance with Cato PoPs
  • Works well for branch offices with physical network requirements

Cons:

  • Smaller than the top 3
  • Requires Cato hardware or virtual appliances for branch
  • Less ecosystem integration than Zscaler or Cloudflare

Best fit: mid-to-large enterprise with branch offices, wanting one vendor for all network security.

Tailscale

Pros:

  • Developer-favorite mesh VPN with ZTNA policies (ACLs)
  • WireGuard-based, low-latency direct peer connections
  • Simple to deploy, self-service for engineering teams
  • Pricing scales well for small-to-mid teams
  • Device posture via integrations

Cons:

  • Not a full ZTNA platform. No built-in SWG, CASB, or DLP
  • "Mesh" model may not fit all enterprise architectures
  • Compliance certifications trail the big vendors

Best fit: engineering-heavy organizations, startups to mid-market, companies that want to replace VPN for internal tooling access without a big-enterprise SSE contract.

Twingate

Similar to Tailscale in positioning. Split-tunnel ZTNA focused on replacing VPN for specific resource access. Good UX, clean Terraform, reasonable pricing.

Google BeyondCorp Enterprise

Google's commercialized BeyondCorp. Strong if you're already a Google Workspace shop. Integrates natively with Chrome, Google Identity, Google Cloud. Less mature for non-Google environments.

The implementation roadmap

Roughly the 18-month rollout we guide clients through during Zero Trust engagements.

Phase 0: baseline (months 0-2)

  • Asset inventory. You can't protect what you don't know exists
  • Identity inventory. Who are your users, what groups exist, what apps do they access
  • Network inventory. What's on the corporate network, what's in which VPC, what's in which SaaS
  • Existing security stack audit. IdP, MFA coverage, EDR coverage, SIEM coverage, VPN usage
  • Executive sponsorship. ZTA is a multi-year program. Without CISO + CIO + CEO buy-in it stalls

Phase 1: identity foundation (months 2-4)

  • SSO to every application (target 100%, practical target 90%+ in phase 1)
  • MFA required for every user, every application (phishing-resistant preferred)
  • SCIM provisioning for top 20 applications
  • Privileged account inventory and enforcement of admin MFA
  • Legacy authentication (basic auth, NTLM, SMBv1) killed everywhere possible

Phase 2: device trust (months 4-7)

  • MDM enrollment mandatory for all endpoints accessing corporate resources
  • EDR deployed on all endpoints with posture reporting
  • Device certificates issued via cloud PKI
  • Posture checks integrated into IdP access decisions (device must be MDM-enrolled, disk-encrypted, EDR-reporting, OS patched)

Phase 3: ZTNA deployment (months 5-10)

  • Pick your ZTNA vendor
  • Start with a single application (internal admin tool, internal wiki)
  • Validate user experience, policy engine, logging
  • Gradually migrate applications. Prioritize crown jewels first, legacy apps last
  • VPN decommission target: end of phase 3

Phase 4: micro-segmentation (months 8-14)

  • Start with the server estate, not user workstations
  • Service-to-service mTLS via service mesh for containerized workloads
  • VPC/VLAN segmentation for legacy infrastructure
  • Identity-based network policies (Istio AuthorizationPolicies, Cilium NetworkPolicies, AWS VPC Lattice)

Phase 5: continuous verification and response (months 10-18)

  • UEBA rules for lateral movement, impossible travel, privilege escalation
  • Automated session revocation workflows (user reported phishing → revoke all sessions → force reauth)
  • SOAR playbooks for common Zero Trust violations
  • Regular red team exercises testing the Zero Trust controls

The mistakes that turn Zero Trust into theater

The VPN with identity in front. You bought ZTNA, deployed the client. But configured it to route all traffic to the corporate network via a single gateway with broad access. That's a VPN with an IdP. Real Zero Trust requires per-application policies, not "all corporate net."

MFA everywhere except the legacy app. The exception always becomes the attack path. Every app without MFA is an MFA bypass.

Device trust means MDM enrolled. And nothing else. If your posture check is "is the device in Jamf" you haven't checked device posture. Check EDR running, OS patched, disk encrypted, no jailbreak, last seen within 24 hours.

Network segmentation via VLANs only. VLANs without identity-aware policies allow any compromised device in the VLAN to pivot to any other. Layer 2 segmentation with no identity layer on top isn't Zero Trust.

No continuous verification. User authenticates at 9am. At 2pm the session is still trusted. At 4pm the laptop is stolen. The session should have continuous re-verification. Device still healthy, user behavior not anomalous, access still justified. Most deployments skip this.

No logging or response automation. Zero Trust produces a massive log volume. Without a SIEM ingesting and a SOAR responding, the logs are expensive storage.

"We deployed ZTNA for remote users, the office is still flat". This isn't Zero Trust, this is segmented remote access. Office users should go through the same authentication and authorization as remote users.

Third parties still get VPN. Vendors, contractors, MSPs still have VPN credentials because "ZTNA is too complex for them." No. Especially third parties need ZTNA. Target got breached through an HVAC vendor on a VPN in 2013. The same pattern repeats every year.

Service-to-service still uses network trust. Your apps talk to each other inside the VPC with no mTLS, no service identity, no authorization. If an attacker gets RCE on one service they get lateral movement everywhere. Service mesh or VPC Lattice with explicit service identity is required.

Break-glass accounts without controls. "we've to have an account for emergencies." Fine. That account needs hardware MFA only, daily log review, rotating credentials stored in a physical safe with signed checkout. And automated alerts on any use.

Zero Trust and compliance

Zero Trust maps cleanly to several compliance frameworks:

  • NIST 800-171 / CMMC Level 2. Access control, identification and authentication, audit and accountability
  • NYDFS 23 NYCRR 500. MFA requirements, privileged access controls, asset inventory, continuous monitoring
  • PCI DSS 4.0. Segmentation, identity verification, MFA on administrative access
  • HIPAA Security Rule. Access controls, audit controls, transmission security
  • SOC 2. CC6.1 (logical access), CC6.6 (network security), CC7.2 (monitoring)
  • ISO 27001:2022. A.5.15 (access control), A.8.2 (privileged access), A.8.16 (monitoring activities)

The CISA Zero Trust Maturity Model

CISA published the Zero Trust Maturity Model v2.0 in 2023, updated 2025. Five pillars:

  1. Identity. From traditional (password + on-prem MFA) to optimal (continuous validation with risk signals, phishing-resistant MFA, automated identity lifecycle)
  2. Devices. From traditional (manual inventory) to optimal (continuous verification, automatic enforcement based on posture)
  3. Networks. From traditional (segmented with VLANs) to optimal (fully encrypted, dynamic least-privilege access, micro-segmentation per-workload)
  4. Applications and Workloads. From traditional (perimeter-protected) to optimal (continuous authorization, immutable infrastructure, integrated threat protection)
  5. Data. From traditional (static classifications) to optimal (automated classification, encryption in use, DLP enforcement, continuous discovery)

Most organizations in 2026 sit at "Initial" or "Advanced" across most pillars. "Optimal" requires continuous investment and significant tooling maturity.

Resources

  • NIST SP 800-207 Zero Trust Architecture: https://csrc.nist.gov/publications/detail/sp/800-207/final
  • CISA Zero Trust Maturity Model: https://www.cisa.gov/zero-trust-maturity-model
  • Google BeyondCorp paper: https://research.google/pubs/beyondcorp-a-new-approach-to-enterprise-security/
  • OMB M-22-09: https://www.whitehouse.gov/wp-content/uploads/2022/01/M-22-09.pdf
  • DoD Zero Trust Strategy: https://dodcio.defense.gov/Portals/0/Documents/Library/DoD-ZTStrategy.pdf
  • Gartner Market Guide for ZTNA (2024-2025)
  • Forrester ZTNA Wave

Hire Valtik Studios

Zero Trust implementations fail when they become point-solution deployments instead of architectural programs. We run Zero Trust assessments against the CISA Maturity Model, produce a phased implementation roadmap mapped to your existing tooling. And validate the controls through penetration testing that specifically targets lateral movement and authentication bypasses. If you're a federal contractor, NYDFS-regulated financial, or defense-adjacent, we align the Zero Trust work with the compliance framework so you get audit credit for the security program.

Reach us at valtikstudios.com.

zero trustZTNABeyondCorpidentitynetwork securityCISO

Want us to check your Zero Trust setup?

Our scanner detects this exact misconfiguration. plus dozens more across 38 platforms. Free website check available, no commitment required.

Get new research in your inbox
No spam. No newsletter filler. Only new posts as they publish.