Zero Trust · Saudi Arabia

Zero Trust is not a product, and no one can sell you one

It is a position: every access request is authorised on its own merits, not on where it came from. Crux implements that position in the order that reduces exploitable risk fastest — which is rarely the order the diagrams suggest.

Five pillars Identity, devices, networks, applications and workloads, data — CISA's maturity model
NIST SP 800-207 The architecture standard the pillars sit beneath
12–16 weeks To a foundation covering privileged access, phishing-resistant MFA and first segmentation
NCA aligned Control mapping produced as a deliverable, against your applicable ECC version
01 — Beyond the slogan

"Never trust, always verify" is a slogan. Here is the actual definition.

Zero Trust means an access decision is made per request, evaluating who is asking, what device they are asking from, what they want, and the surrounding context — instead of granting access because the request originated inside the network. NIST SP 800-207 sets out the architecture. CISA's maturity model organises implementation into five pillars, which is where the familiar diagram comes from.

The reason it matters is narrower than the marketing suggests. Traditional networks are hard on the outside and permissive within, so a single compromised laptop becomes access to everything that laptop could reach. Nearly every significant incident follows this shape: initial access through phishing or an exposed service, then movement.

Zero Trust is the discipline of making the second step expensive. It does not prevent initial compromise, and any vendor implying otherwise is overselling. What it does is shrink what an attacker can reach once inside — which is the difference between an incident and a catastrophe.

02 — Blast radius

The same intrusion, in two different networks

A finance user opens an attachment. Their credentials and workstation are compromised. What happens next depends entirely on architecture, not on the attachment.

Flat network with perimeter trust
  1. Attacker inherits the user's network position and can reach anything routable
  2. Scans internally without obstruction, mapping file shares, databases and management interfaces
  3. Finds a service account with a stored password and standing administrative rights
  4. Moves to a domain controller and escalates to enterprise-wide privilege
  5. Stages exfiltration and deploys ransomware across everything reachable
  6. Detection arrives with the ransom note
Zero Trust architecture
  1. Attacker has credentials but the device fails posture checks on the next request
  2. Lateral scanning is blocked by segmentation; unrelated workloads are not reachable
  3. No standing administrative privilege exists to find; elevation is time-bound and approved
  4. Anomalous access patterns surface in monitoring because normal is defined
  5. Blast radius is limited to what that one identity was explicitly entitled to
  6. Detection arrives while the incident is still one workstation

Neither column prevents the attachment being opened. That is the point: Zero Trust is a containment architecture, not a prevention one.

03 — The five pillars

What each pillar means in an actual Saudi enterprise

CISA's five pillars, described as work rather than as a diagram. Most organisations are strong in one or two and largely absent in the rest, which is what an assessment establishes.

  • 01 Identity

    The pillar everything else depends on. Phishing-resistant MFA rather than SMS codes, conditional access evaluating context, joiner-mover-leaver processes that actually revoke, and Nafath where verified national identity is required. Weak identity makes the remaining four pillars decorative.

    Phishing-resistant MFAConditional accessNafathLifecycle management
  • 02 Devices

    Access granted to a known, compliant endpoint rather than to a credential alone. Inventory first — organisations are routinely surprised by what is connecting — then posture checks, patch state and certificate-based identity, including a workable position for contractor and BYOD devices.

    InventoryPosture checksMDMBYOD policy
  • 03 Networks

    Segmentation so that reaching one system does not mean reaching all of them. Start coarse: separate crown-jewel systems, management interfaces and OT environments from general corporate traffic, then refine. Micro-segmentation attempted at full granularity on day one usually stalls.

    SegmentationEast-west inspectionOT/IT separationZTNA
  • 04 Applications and workloads

    Access enforced at the application layer, not merely at the network edge. Discovery of unsanctioned services, per-application authorisation, API protection, and workload identity so services authenticate to each other rather than trusting a source address.

    Per-app accessShadow IT discoveryAPI securityWorkload identity
  • 05 Data

    The pillar most often deferred and most directly tied to consequence. Classification so controls can be proportionate, encryption, loss prevention on the paths that matter, and access logging sufficient to answer what was taken — which is also a PDPL requirement.

    ClassificationEncryptionDLPAccess logging
04 — Maturity

Where organisations actually sit

Assessment output, simplified. Most Saudi enterprises we assess land in "traditional" or "initial" across most pillars, with identity usually furthest along because MFA projects happened first.

Pillar Traditional Initial Advanced
IdentityPasswords, shared admin accounts, manual provisioningMFA on most users, some conditional accessPhishing-resistant MFA everywhere, no standing privilege, automated lifecycle
DevicesUnknown inventory, unmanaged endpoints on the networkManaged estate, patch reporting existsAccess conditional on live posture; non-compliant devices cannot connect
NetworksFlat internal network, VPN grants broad accessVLAN separation, some firewall rules between zonesWorkload-level segmentation, ZTNA for remote and third-party access
ApplicationsNetwork access implies application accessSSO for major applicationsPer-application authorisation, workload identity, unsanctioned services discovered
DataUnclassified, access poorly recordedClassification defined on paperClassification enforced technically, DLP on key paths, access fully logged

Mapping to NCA ECC control identifiers is produced against the ECC version applicable to your organisation and sector, since requirements have been revised over time.

05 — Sequencing

What to do first, and why it is not segmentation

Segmentation is usually attempted first because it feels like architecture and produces impressive diagrams. It is also slow, disruptive, and dependent on an application dependency map most organisations do not have.

Privileged access is the better first move. Standing administrative privilege is what converts a compromised workstation into an enterprise incident, and removing it is faster, cheaper and less disruptive than any network project.

  • Weeks 1–4: privileged accounts. Find every one, including service accounts with passwords in scripts. Vault credentials, remove standing privilege, make elevation time-bound and recorded.
  • Weeks 1–6: phishing-resistant MFA. Administrators and remote access first. SMS and push-approval are no longer adequate against the phishing techniques in current use.
  • Weeks 2–8: know what exists. Device and application inventory, running in parallel. Controls cannot be applied to systems nobody has listed.
  • Weeks 6–14: coarse segmentation. Separate crown jewels, management interfaces and OT from general traffic. Refine later; the first cut delivers most of the containment benefit.
  • Weeks 10–16: third-party and remote access. Move contractors and vendors to per-application ZTNA. Third-party access is a recurring initial vector and is usually the least governed.
  • Ongoing: data classification. Slow and organisational rather than technical, so start early and expect it to run alongside everything else.
06 — AI in Zero Trust

Useful once identity is in order. Noise before that.

Behavioural analytics on top of weak identity hygiene produces alerts nobody can action, because "normal" cannot be defined when accounts are shared and privilege is permanent. Applied after the foundations, the same techniques become genuinely valuable.

Behavioural baselining

Learning what normal access looks like per identity and per workload, so deviation becomes a signal. This is what makes "continuous verification" more than a phrase — context that changes mid-session can trigger re-authentication.

UEBA · Session risk

Access anomaly detection

Impossible travel, unusual privilege use, access to systems an identity has never touched. Individually weak signals; combined, a reliable early indicator of credential compromise.

Credential misuse · Early warning

Policy recommendation

Segmentation is hard because nobody knows which systems legitimately talk to each other. Traffic analysis proposes policies from observed behaviour instead of from a whiteboard, which is often the difference between a stalled project and a delivered one.

Traffic analysis · Policy mining

Alert triage

Zero Trust generates more telemetry, not less. Clustering related alerts and suppressing known-benign patterns is what keeps a small security team able to respond to the ones that matter.

Correlation · Noise reduction

Entitlement review

Access accumulates as people change roles and nobody removes the old permissions. Models comparing entitlements against peers and actual usage surface excess privilege far faster than an annual manual review.

Least privilege · Access review

Third-party risk monitoring

Vendor and contractor access is a recurring initial vector. Continuous monitoring of that access, rather than an onboarding questionnaire, is where the actual risk sits.

Vendor access · Continuous
07 — Delivery

Risk down in the first quarter, not the second year

Zero Trust programmes that deliver nothing measurable for eighteen months lose sponsorship before they finish. Each phase here is scoped to reduce a named risk.

  1. Maturity assessment

    Current state across the five pillars, with the specific paths an attacker would take through your environment. Mapped to your applicable NCA ECC version and sector requirements.

    3 weeks
  2. Privileged access

    Discovery of privileged and service accounts, vaulting, removal of standing privilege, time-bound elevation with recording. The single largest risk reduction available.

    4–6 weeks
  3. Identity hardening

    Phishing-resistant MFA, conditional access policies, and lifecycle processes that revoke reliably when people leave or change role.

    4–6 weeks
  4. Segmentation

    Traffic analysis to establish real dependencies, then coarse segmentation of crown jewels, management planes and OT, with policy refined against observed behaviour.

    8–12 weeks
  5. ZTNA rollout

    Per-application access for remote and third-party users, running alongside the VPN while legacy cases migrate.

    6–8 weeks
  6. Data and continuous improvement

    Classification, DLP on the paths that matter, and behavioural monitoring introduced once identity and segmentation make the signals meaningful.

    Ongoing

A foundation lands in 12 to 16 weeks. Maturity across all five pillars is a 12 to 24 month programme for a large enterprise — but exploitable risk should fall well before then, and if it does not, the sequencing is wrong.

08 — Questions

Answered plainly

What is Zero Trust, beyond the slogan?

Access decisions made per request, based on who is asking, what device they are asking from, and the context — rather than granted because the request came from inside the network. NIST SP 800-207 sets out the architecture; CISA's maturity model organises it into five pillars: identity, devices, networks, applications and workloads, and data.

Is Zero Trust a product we can buy?

No, and vendors selling it as one are selling a component. It is an architectural position implemented through identity, endpoint, network and data controls working together. Products are necessary, but sequencing and the policy model determine whether the outcome is real.

Where should we start?

Privileged accounts and phishing-resistant MFA, then removing standing administrative privilege. These cut the most exploited path faster and more cheaply than segmentation, which is usually attempted first because it feels like architecture. Inventory runs alongside — controls cannot be applied to systems nobody has listed.

How does this relate to NCA ECC?

Zero Trust controls map onto ECC subdomains covering identity and access management, network security management and system protection. The mapping is made against the ECC version applicable to your organisation and sector, since identifiers and requirements have been revised. We produce that mapping as a deliverable rather than asserting compliance generically.

Does Zero Trust replace our VPN?

Usually over time. ZTNA grants per-application access after verifying identity and device posture, rather than placing a device on the network. Migration is gradual: high-risk and third-party access moves first while the VPN continues serving legacy cases.

How long does implementation take?

A foundation covering privileged access, phishing-resistant MFA and initial segmentation typically takes 12 to 16 weeks. Maturity across all five pillars is a 12 to 24 month programme for a large enterprise. A good programme reduces exploitable risk in the first quarter, not the last.

Start here

Find out how far an attacker would get from one compromised laptop

A three-week assessment maps your maturity across the five pillars, traces the paths an attacker would actually take through your environment, and returns a sequenced plan ranked by risk reduction rather than by tidiness.