Zero Trust, Defense in Depth & Defender for Cloud

AZ-900 AZ-900 · Describe Azure architecture and services

By Abu Anns · Last updated 20 November 2026Exam version: AZ-900 (skills measured as of July 20, 2026)

Zero Trust: never trust, always verify

The traditional security model trusted everything inside the corporate network — once you were through the firewall, you were trusted. That model breaks in the cloud, where users work from anywhere and applications live on the internet. Zero Trust replaces it with a simple rule: never trust by default, always verify. Every access request is authenticated, authorised, and encrypted, regardless of where it comes from — inside or outside the network. Location grants no trust; identity, device health, and risk do.

Microsoft describes Zero Trust through three principles, and the exam tests all three:

Zero Trust is a strategy, not a product — no single service "is" Zero Trust. Entra ID (identity verification), Conditional Access (risk-based decisions), MFA (strong authentication), RBAC (least privilege), and Defender for Cloud (threat detection) are the Azure tools that implement it. If a question describes verifying every request regardless of network location, the answer is Zero Trust.

Defense in depth: layers, not a single wall

Defense in depth is the strategy of protecting a system with multiple layers of security, so that if one layer fails, the next one still stands. A single firewall can be bypassed; seven layers rarely are. The exam expects you to describe the layers, from the outside in:

LayerWhat it protectsExample controls
Physical securityThe datacenter buildingGuards, locks, cameras, biometric access
Identity and accessWho can get inEntra ID, MFA, Conditional Access, RBAC
PerimeterThe network edgeDDoS protection, firewalls
NetworkTraffic between resourcesNetwork security groups, private endpoints, segmentation
ComputeServers and virtual machinesEndpoint protection, patching, disk encryption
ApplicationThe software itselfSecure code, application firewalls, secrets management
DataThe informationEncryption at rest and in transit, access controls

The layers work together: an attacker who gets past the perimeter firewall still faces network segmentation; one who reaches a VM still faces identity checks and encryption; one who reaches the data still faces encryption. The exam tests defense in depth with scenarios about "multiple layers" or with questions asking which layer a control belongs to — MFA belongs to identity and access, network security groups to the network layer, encryption to the data layer.

Microsoft Defender for Cloud

Microsoft Defender for Cloud is Azure's cloud security posture management and threat-protection service. It does two jobs. First, it assesses the security of your Azure (and connected on-premises and multi-cloud) resources and gives you a secure score — a percentage showing how closely your configuration matches security best practices — plus prioritised recommendations for improving it (enable MFA, encrypt disks, close open management ports). Second, it detects threats: it monitors for suspicious activity, such as a virtual machine communicating with a known malicious address, and raises security alerts with recommended actions.

Defender for Cloud has a free tier (secure score, recommendations, and the asset inventory) and paid Defender plans per resource type (servers, storage, databases, and more) that add threat detection. The exam does not test pricing tiers; it tests purpose: Defender for Cloud tells you how secure your environment is and warns you about threats. Do not confuse it with Microsoft Sentinel (a SIEM for collecting and analysing security events across an organisation — not in the AZ-900 objectives) or with Microsoft Defender for Endpoint (protection for individual devices). For AZ-900, Defender for Cloud is the cloud security posture and alerting service.

Worked example: one attack, three concepts

An attacker phishes an employee's password. Under Zero Trust, the password alone is not enough: Conditional Access sees a sign-in from an unusual location on an unknown device and requires MFA, which the attacker cannot complete. Suppose the attacker somehow gets in anyway. Defense in depth limits the damage: network segmentation keeps them away from the database subnet, RBAC least privilege means the stolen account can read only one application, and encryption means any data they reach is unreadable without keys. Meanwhile Defender for Cloud detects the anomalous sign-in pattern and raises an alert, and the security team disables the account. Three exam concepts — Zero Trust, defense in depth, Defender for Cloud — describe three lines of the same defence.

How the security topics connect

The identity guide and this guide are two halves of one domain area. Identity (Entra ID, MFA, Conditional Access, RBAC) answers "who are you and what can you do?"; this guide answers "what is the strategy, and how do we know we are secure?" In the exam, a question about verifying every request is Zero Trust; a question about layered protection is defense in depth; a question about a secure score or security recommendations is Defender for Cloud; and a question about who can do what is RBAC. Keep those four triggers straight and the security questions answer themselves.

The exam's security checklist

Before the exam, check you can answer these five questions without hesitation. What are the three Zero Trust principles, and what does each one mean in practice? Which defense-in-depth layer does each of these belong to: MFA, a network security group, disk encryption, a firewall, a security guard? What does Defender for Cloud give you — a secure score and recommendations — and what does it detect? What is the difference between Owner and Contributor, and why does least privilege matter? And what does Conditional Access use as signals — user, device, location, risk — and what controls can it enforce? If you can answer all five in your own words, the security portion of the exam is covered. The most common mistake is treating these as separate facts to memorise; they are one story — identity proves who you are, Zero Trust decides whether to believe it, defense in depth limits the damage if the answer is wrong, and Defender for Cloud watches the whole thing and tells you what to fix.

Advertisement

Exam traps to watch for

Practice questions

Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.

Q1.A company adopts a security model in which every access request is verified based on identity, device, and location — whether the user is in the office or working remotely. No request is trusted just because it comes from inside the network. Which model is this?

Report an error

Q2.Which of the following is one of the three principles of Zero Trust?

Report an error

Q3.A company protects its systems with a firewall at the network edge, network segmentation, endpoint protection on servers, and encryption of its data, so that if one control fails, the others still protect the system. Which strategy is the company using?

Report an error

Q4.In the defense-in-depth model, which layer does multifactor authentication (MFA) belong to?

Report an error

Q5.A security team wants a single view of how securely its Azure resources are configured, with a score and a prioritised list of improvements. Which service provides this?

Report an error

Q6.Defender for Cloud raises an alert that a virtual machine is communicating with a known malicious IP address. Which Defender for Cloud capability produced this alert?

Report an error

Q7.According to the Zero Trust principle of least privilege, what access should a user receive?

Report an error

Q8.A company encrypts its databases and storage so that data is unreadable without the keys, even if an attacker gains access to the underlying files. Which defense-in-depth layer is this control part of?

Report an error