The SC-100 blueprint is split across 4 domains. Weights are the official share of the exam; see the official exam guide for the authoritative breakdown.
What you must be able to do. Given a scenario about resilience, framework alignment, or where to begin, choose the design that matches the Microsoft reference architecture, benchmark, or adoption framework the requirement points to, and that survives a fully compromised privileged identity.
In one sentenceThe strategy domain: aligning designs to MCRA, MCSB, CAF, and WAF, sequencing work with the Rapid Modernization Plan, and building ransomware resilience that holds even when an admin is compromised.
Recall check: answer these from memory first
- A ransomware operator already holds Global Administrator and intends to delete every backup before detonating. Which single backup design property defeats this, and why does cross-region replication not?
- Which Microsoft framework answers each: reference architecture diagrams and capability mapping, a prescriptive control baseline, the cloud operating model and landing zones, and the security pillar with its trade-offs?
- The Rapid Modernization Plan sequences which workstream first, and why does Microsoft place it ahead of network segmentation and endpoint onboarding?
What it tests. Designing a security strategy from Microsoft's own best-practice guidance rather than ad hoc preference. Building a resiliency strategy for ransomware and other attacks, where backups and the identity foundation must survive an attacker who already holds privileged rights; aligning designs to the Microsoft Cybersecurity Reference Architectures (MCRA) and the Microsoft cloud security benchmark (MCSB) for reference patterns and control baselines; aligning to the Cloud Adoption Framework for Azure (CAF) for landing zones, governance, and DevSecOps, and to the Azure Well-Architected Framework (WAF) for the security pillar and trade-offs; and sequencing the highest-leverage work with the Rapid Modernization Plan, which leads with privileged access and identity.
How to study it. Make assume-breach your reflex, because the resilience questions are decided by it. When a scenario names a compromised Global Administrator or a ransomware operator targeting backups, the answer is the control the attacker cannot defeat with the access they hold: immutable, time-locked backup storage, and a dedicated isolated recovery path for the directory itself. Learn what each framework is for so you map the requirement to the right one: MCRA for reference architecture diagrams and capability mapping, MCSB for prescriptive control baselines, CAF for the operating model, landing zones, and shift-left DevSecOps such as policy-as-code in the pipeline, and WAF for the security pillar and its trade-offs against cost and operations. Fix the Rapid Modernization Plan ordering: it sequences privileged access and identity first as the highest-leverage early risk reduction, ahead of network segmentation, data protection, or endpoint onboarding.
Easy to confuse
- Immutable time-locked backups versus cross-region replication. Immutability with a retention lock makes restore points undeletable by any role, including a compromised Global Administrator, which is the ransomware threat; cross-region replication only survives a regional outage and still lets a privileged attacker delete both copies. Resilience against a compromised admin needs immutability, not geographic redundancy.
- MCRA versus MCSB. The Microsoft Cybersecurity Reference Architectures are diagram-led reference patterns that map capabilities across the estate; the Microsoft cloud security benchmark is a prescriptive set of control recommendations measured in Defender for Cloud. Reach for MCRA when the scenario asks how the pieces fit together, MCSB when it asks which controls to enforce and measure.
- Cloud Adoption Framework versus Well-Architected Framework. CAF guides the cloud operating model, landing zones, governance, and adoption lifecycle including DevSecOps; WAF evaluates a workload against five pillars and the security-versus-cost-and-operations trade-offs within it. CAF for how the organisation adopts and governs the cloud, WAF for how a specific workload is engineered.
Worked example from the SC-100 bank
lock_openFree sampleDesign Solutions that Align with Security Best Practices and Prioritieshard
A government agency wants its ransomware recovery design to guarantee that, after an attack, it can restore not just files but a trustworthy version of its core identity foundation, so it can re-establish authentication and authorisation before applications come back. A review finds the agency backs up application data and virtual machines but has no separate, validated recovery path for the directory that everything else depends on. Which design addition most directly closes this gap?
- AEstablish a dedicated, isolated recovery path for the identity foundation with secured backups of directory state, so a trustworthy directory can be rebuilt before applications are restored.check_circle Correct
- BAdd application-consistent snapshots of every virtual machine more frequently so that the servers hosting the directory can be rolled back quickly alongside the rest of the estate.
- CReplicate the directory to a warm standby in a second region so that if the primary is encrypted the agency can fail authentication over to the continuously synchronised copy.
- DTighten Conditional Access on all directory administrator accounts so that a compromised admin cannot make the changes that would damage the directory during an attack.
Resilient recovery needs a dedicated isolated path with secured directory backups so a trustworthy identity foundation can be rebuilt before applications are restored. Applications cannot return until authentication and authorisation are trustworthy, so the identity foundation needs its own validated recovery path rather than ordinary machine backups or live replication that may carry the compromise forward. A dedicated, isolated recovery path with secured directory-state backups lets the organisation rebuild a clean directory first, which frequent snapshots, warm standbys, or tighter admin access cannot guarantee.
Why A is correct: A dedicated isolated recovery path with secured backups of the directory lets the agency rebuild a trustworthy identity foundation first, which is the prerequisite for restoring authentication and authorisation before applications, exactly as the requirement states.
Why B is wrong: More frequent machine snapshots speed server rollback and seem to cover the directory hosts, but restoring infected or tampered directory state from ordinary snapshots can reintroduce the compromise, so it does not provide the validated trustworthy directory recovery required.
Why C is wrong: A synchronised standby aids availability and is tempting for fast failover, but live replication faithfully copies a compromised or encrypted directory to the standby, so the agency fails over into the same tainted state rather than a trustworthy one.
Why D is wrong: Hardening admin access reduces the chance of directory damage and is worthwhile, but it is a preventive control that provides no recovery path, so once the directory is destroyed the agency still has no validated way to rebuild it.
What you must be able to do. Given a scenario about detection and response, identity and access, privileged access, or regulatory compliance, choose the design that enforces the requirement at grant time or in near real time and produces auditable accountability, using the right Microsoft operations, identity, and Purview capability.
In one sentenceThe largest domain: designing security operations, identity and access governance, privileged access, and compliance so enforcement is preventive and near real time rather than detective and after the fact.
Recall check: answer these from memory first
- A revoked or risky user must lose access mid-session within minutes on Exchange and SharePoint Online. Which capability enforces this, and why do shorter token lifetimes and sign-in frequency fall short?
- Which design prevents a user from ever holding two incompatible privileged roles at once, and why is recurring access review or manager approval not preventive?
- An auditor needs control implementation, named owners, and stored evidence for a standard with no built-in template. Which Microsoft tool manages this, and why not Azure Policy or Defender for Cloud?
What it tests. Designing the operations, identity, and compliance backbone. Security operations across detection, response, logging, SOAR, and threat-detection coverage with Microsoft Sentinel and the Microsoft Defender XDR suite; identity and access management across SaaS, PaaS, IaaS, hybrid, and multicloud with Microsoft Entra ID, including near-real-time enforcement such as continuous access evaluation and risk-based Conditional Access; privileged access using the enterprise access model, with Privileged Identity Management for just-in-time activation and entitlement management for access packages and separation of duties; service-account hardening such as group managed service accounts in Active Directory Domain Services; and translating regulatory requirements into security controls with Microsoft Purview Compliance Manager assessments, ownership, and evidence.
How to study it. Drill three distinctions the domain returns to again and again. First, near-real-time enforcement versus fixed timers: continuous access evaluation pushes revocation and critical events to supporting services in minutes, which a shorter token lifetime or tighter sign-in frequency cannot match because they only act at the next interval. Second, preventive versus detective governance: an entitlement management separation-of-duties rule blocks a request for an incompatible access package at grant time, whereas manager approval, recurring access reviews, or just-in-time activation only catch or limit the toxic combination after the fact. Third, the right home for control management: a custom Microsoft Purview Compliance Manager assessment decomposes an arbitrary standard into improvement actions with owners and evidence, which resource-scoped tools like Azure Policy or Defender for Cloud cannot. For service accounts, fix group managed service accounts as the design that lets Active Directory generate, rotate, and scope passwords so no administrator ever knows them.
Easy to confuse
- Continuous access evaluation versus shorter token lifetime or sign-in frequency. Continuous access evaluation has Microsoft Entra ID push critical events such as account disablement, password change, and elevated risk to supporting services, which reject an already issued token within minutes; a shorter token lifetime or sign-in frequency only re-checks at the next fixed interval, leaving a window open. Near-real-time revocation needs CAE, not a tighter timer.
- Entitlement management separation of duties versus access reviews. An entitlement management separation-of-duties rule blocks the request for an incompatible access package at grant time, so the toxic combination never forms; recurring access reviews only detect and remove an accumulated conflict later, after the risk has existed. Preventive enforcement needs the grant-time rule, detective cleanup is too late.
- Group managed service accounts versus manual password rotation or Local System. Group managed service accounts let Active Directory generate, rotate, and scope a long random password to authorised hosts so no administrator ever knows or stores it; manual rotation still leaves an admin-known shared secret, and moving to Local System changes the security context rather than solving credential management. Eliminating the admin-known shared password needs gMSAs.
- Compliance Manager assessment versus Azure Policy or Defender for Cloud. A Microsoft Purview Compliance Manager assessment tracks control implementation, assigns improvement actions to owners, and stores audit evidence for any standard, including custom ones; Azure Policy and Defender for Cloud measure resource configuration against technical baselines but do not manage ownership and evidence for an arbitrary regulatory standard. Auditable control accountability lives in Compliance Manager.
Worked example from the SC-100 bank
lock_openFree sampleDesign Security Operations, Identity, and Compliance Capabilitieshard
A bank's separation-of-duties policy states that no single person may simultaneously hold both the role that approves payment batches and the role that releases funds, because together they enable undetected fraud. The architect is designing how staff request these privileged roles and must ensure the access governance design itself blocks anyone from obtaining the second role while they hold the first. Which design best enforces this incompatible-access rule?
- ARequire manager approval on each role request so an approver can manually notice when a requester already holds the conflicting role and decline the request.
- BPlace both roles under Privileged Identity Management so each must be activated just in time, limiting how long either role is held at once.
- CRun recurring access reviews across both roles so reviewers can detect and remove anyone who has accumulated the conflicting combination over time.
- DDefine both roles as Microsoft Entra entitlement management access packages with a separation-of-duties rule that prevents requesting one package while assigned the incompatible package.check_circle Correct
Use entitlement management separation-of-duties rules between access packages to prevent a user from obtaining incompatible privileged roles. Separation of duties is enforced preventively when the access governance layer encodes which entitlements are mutually exclusive; entitlement management separation-of-duties rules block a request for one access package while the requester holds the incompatible package, stopping the toxic combination at grant time.
Why A is wrong: It is tempting because approval adds a human gate, but it is wrong because it depends on an approver remembering and checking conflicts, which is error-prone and does not systematically enforce the incompatible-access rule.
Why B is wrong: It is tempting because PIM limits standing time, but it is wrong because a user could still activate both eligible roles simultaneously, since PIM does not encode mutual exclusivity between incompatible roles.
Why C is wrong: It is tempting because reviews catch accumulated access, but it is wrong because detection happens only at review intervals after the conflict exists, rather than preventing the incompatible combination from being granted in the first place.
Why D is correct: It is correct because entitlement management supports separation-of-duties constraints between access packages, automatically blocking a request for one package while the requester holds the incompatible one, enforcing the rule at request time.
What you must be able to do. Given a scenario about posture, endpoints, workloads, or network edge across hybrid and multicloud, choose the design that matches protection to the exact surface, distinguishing configuration posture from runtime detection and selecting the right Defender plan or Security Service Edge pattern.
In one sentenceThe infrastructure domain: posture management across hybrid and multicloud, the correct Defender workload plan for each surface from endpoints to shipped IoT, and Security Service Edge for agentless network onboarding.
Recall check: answer these from memory first
- Defender for Cloud already gives posture recommendations, but crypto-mining and a suspicious shell ran undetected on the servers. Which addition gives runtime detection, and why does raising Secure Score not?
- Match the Defender design to the surface: a managed laptop, an Arc-enabled server, a network of devices you only monitor, and a wearable you manufacture and ship into the field.
- A branch with agentless point-of-sale terminals must route its traffic through the cloud secure web gateway by egress. Which Global Secure Access design fits, and why not a per-device VPN or agent?
What it tests. Matching infrastructure protection to the surface it must defend. Posture management across hybrid and multicloud with Microsoft Defender for Cloud, Secure Score, the Microsoft cloud security benchmark, and multicloud connectors; securing server and client endpoints across platforms, mobile, IoT, and operational technology, including Windows LAPS for local administrator password rotation and Microsoft Defender for Endpoint; securing SaaS, PaaS, and IaaS including containers and workloads, with the Defender for Cloud workload protection plans such as Defender for Servers for runtime threat detection; protecting devices an organisation builds and ships with the Defender for IoT device builder micro-agent; and evaluating network security and Security Service Edge with Global Secure Access, Microsoft Entra Internet Access, and Microsoft Entra Private Access.
How to study it. Lock two splits that decide most of this domain. First, posture versus runtime: posture management scores how resources are configured, but it cannot see malicious behaviour executing, so when a scenario names crypto-mining or a suspicious shell running on servers, the answer is the Defender for Servers workload protection plan, not a higher Secure Score target. Second, match the Defender plan to the surface: Defender for Endpoint for managed endpoints, Defender for Servers for server workloads, passive Defender for IoT network sensors for monitoring a network you control, and the Defender for IoT device builder micro-agent for embedded firmware on devices you manufacture and ship. For endpoint hardening, fix Windows LAPS as the answer to identical shared local administrator passwords. For network edge, learn that Global Secure Access remote networks build an IPsec tunnel from a site router to the Microsoft Entra Internet Access edge, onboarding a whole site by its egress so agentless devices are covered without per-device software.
Easy to confuse
- Cloud security posture management versus Defender for Servers workload protection. Posture management and Secure Score assess how resources are configured and recommend hardening, but they cannot detect behaviour at runtime; the Defender for Servers plan adds behavioural threat detection that alerts on activity such as crypto-mining and suspicious shells executing on the machine. Detecting active threats needs the workload plan, not a stronger configuration baseline.
- Defender for IoT device builder micro-agent versus Defender for Endpoint or network sensors. The Defender for IoT device builder micro-agent is compiled into constrained embedded firmware so a shipped device reports tampering and anomalies from itself wherever it operates; Defender for Endpoint needs a full endpoint sensor the device cannot run, and passive network sensors only watch a segment you control. Protecting devices you build and ship needs the micro-agent.
- Global Secure Access remote networks versus per-device VPN or agent. Global Secure Access remote networks onboard a whole site by building an IPsec tunnel from the branch router to the Microsoft Entra Internet Access edge and applying traffic forwarding profiles, covering agentless devices by egress; a per-device VPN profile or Global Secure Access client requires software on each device the point-of-sale terminals and printers cannot run. Agentless whole-site onboarding needs remote networks.
Worked example from the SC-100 bank
lock_openFree sampleDesign Security Solutions for Infrastructurehard
A retailer is bringing several branch sites under its Security Service Edge. The branches host point-of-sale terminals and printers that cannot run a client agent, and the design must still route those sites' internet and software-as-a-service traffic through the organisation's cloud edge so the same secure web gateway and tenant controls apply. The architect wants to onboard each site by its egress rather than by installing software on every device. Which design approach meets this requirement?
- ADeploy the Global Secure Access client to a single gateway server at each branch and configure the branch network to route all device traffic through that one machine so the agent forwards it to the edge.
- BProvision an ExpressRoute circuit from each branch to an Azure hub virtual network and route the branches' internet-bound traffic through an Azure Firewall in the hub for inspection and filtering.
- CSet up a point-to-site VPN profile pushed to each terminal so that the devices dial into a VPN gateway, after which their internet traffic is forced through the corporate egress for filtering.
- DConfigure remote networks in Global Secure Access and establish an IPsec tunnel from each branch's edge router to the Microsoft Entra Internet Access edge, applying traffic forwarding profiles to the whole site.check_circle Correct
Onboard whole sites to Microsoft Entra Internet Access with Global Secure Access remote networks over IPsec, covering agentless devices. Global Secure Access supports remote networks, where a branch router builds an IPsec tunnel to the Microsoft Entra Internet Access edge and traffic forwarding profiles govern the site's flows. This covers devices that cannot run the client, applying the secure web gateway and tenant controls by egress rather than per device, which agent-on-a-server, ExpressRoute and point-to-site VPN designs cannot do.
Why A is wrong: Funnelling a branch through one client-hosting server seems to avoid per-device installs, but the Global Secure Access client is a per-user endpoint agent and is not designed to act as a site router for unattended terminals, so this is an unsupported and fragile way to onboard the site.
Why B is wrong: ExpressRoute with hub Azure Firewall inspects branch traffic and is a valid network design, but it is private connectivity into Azure rather than onboarding to the identity-aware Security Service Edge, so it does not bring the sites under the same secure web gateway and tenant-restriction policy plane the requirement names.
Why C is wrong: Point-to-site VPN forces traffic through a corporate egress and appears to centralise filtering, but it requires a VPN client and profile on every terminal, which the requirement explicitly rules out, and it does not deliver the cloud-edge secure web gateway model.
Why D is correct: Remote networks in Global Secure Access connect a site to the Microsoft Entra Internet Access edge over an IPsec tunnel from the branch router, so every device behind it is covered by the secure web gateway and tenant controls without any agent on the terminals, which is exactly the egress-based onboarding required.
What you must be able to do. Given a scenario about Microsoft 365 collaboration, the application lifecycle and workload identities, or data classification and protection, choose the design that places the control at the correct layer and lifecycle stage, scoping least privilege and protecting data without breaking the stated business need.
In one sentenceThe application and data domain: securing Microsoft 365 workloads, applying lifecycle practices such as design-phase threat modelling and least-privilege workload identities, and protecting data with the right layer of masking, encryption, and app-level controls.
Recall check: answer these from memory first
- Design-level flaws such as missing authorisation boundaries must be found before any code is written. Which practice surfaces them, and why can static testing, dynamic testing, or a penetration test not?
- Analysts must see obscured salary and account values in query results without changing stored data or app logic. Which data control fits, and why not Transparent Data Encryption or row-level security?
- A pipeline service principal holds Owner over the whole subscription but only deploys to one resource group. Which change most directly shrinks the blast radius, and why not Key Vault or secret rotation?
What it tests. Placing application and data controls at the right layer and stage. Securing Microsoft 365 productivity and collaboration workloads, including protecting corporate data on unmanaged mobile devices with Microsoft Intune app protection policies; securing applications across the full lifecycle, with threat modelling at the design phase ahead of code, secure pipeline practices, and least-privilege workload identities scoped to the resource and permissions they need; and securing data through classification, encryption, and database protection, including dynamic data masking at the presentation layer, encryption at rest and in transit, and Microsoft Purview information protection and data loss prevention.
How to study it. Train yourself to ask which layer and which lifecycle stage the requirement names, because that is what separates the right answer from the plausible ones. For the lifecycle, threat modelling is the design-phase practice that surfaces authorisation and trust-boundary flaws before code exists, unlike static or dynamic testing and penetration tests that need built or running software. For data, distinguish the layers: dynamic data masking obscures sensitive column values in query results for limited-privilege users without changing stored data or application logic, which encryption at rest and row-level security do not do; choose encryption when the stored value itself must be protected. For workload identities, least privilege is scoping the role assignment to the smallest resource scope and permissions, which shrinks the blast radius of a compromise, whereas network conditions, secret vaulting, or rotation only limit where or how long a credential works. For unmanaged mobile, Intune app protection policies enforce a work PIN, block data transfer to personal apps, and selectively wipe corporate data without enrolling the device.
Easy to confuse
- Threat modelling versus static or dynamic application security testing. Threat modelling reasons about data flows and trust boundaries during the design phase to find architectural and authorisation flaws before code exists; static testing needs committed source and dynamic testing needs a running build, so both act after implementation. Catching design-level flaws before code needs threat modelling.
- Dynamic data masking versus encryption at rest. Dynamic data masking obscures sensitive column values in query results for non-privileged users at the presentation layer while leaving the stored data and application unchanged; Transparent Data Encryption and Always Encrypted protect the stored or transmitted value but do not selectively obscure results for some readers. Showing masked values to limited-privilege analysts needs dynamic data masking.
- Scoping a workload identity versus vaulting or rotating its secret. Scoping the role assignment to the single resource group and the least permissions confines what a compromised identity can do, shrinking the blast radius; storing the secret in Key Vault or rotating it only changes where the credential lives or how long it lasts while leaving its subscription-wide authority intact. Reducing blast radius needs least-privilege scoping.
- Intune app protection policies versus device enrolment with MDM. Intune app protection policies apply a work PIN, block data transfer to personal apps, and selectively wipe only corporate data at the managed-app level on unenrolled personal phones; full mobile device management enrolment manages the whole device, which the scenario's bring-your-own users refuse. Protecting corporate data on unmanaged phones needs app protection policies, not enrolment.
Worked example from the SC-100 bank
lock_openFree sampleDesign Security Solutions for Applications and Datahard
A product team is about to begin building a new customer-facing payments service and asks a security architect when, in the application lifecycle, design-level security flaws such as missing authorisation boundaries and unsafe trust assumptions should be identified. The architect wants the practice that surfaces these flaws before code is written, by reasoning about the system's data flows and trust boundaries. Which practice best meets this requirement?
- APerform threat modeling during the design phase, enumerating data flows and trust boundaries to identify and rank design weaknesses before any code exists.check_circle Correct
- BRun dynamic application security testing against a deployed staging build so that exploitable runtime weaknesses are discovered before the service reaches production traffic.
- CAdd static application security testing to the build pipeline so that insecure coding patterns are flagged automatically every time a developer commits source.
- DSchedule an external penetration test ahead of launch so that an independent team validates the service's defences against realistic attacker techniques first.
Threat modeling is the design-phase practice that reveals architectural and trust-boundary flaws before code exists, unlike testing that runs against built software. Threat modeling reasons about a system's data flows, trust boundaries, and assets to enumerate how it could be attacked while the design is still on paper. Because it precedes implementation, it catches authorisation, trust, and exposure flaws when they are cheapest to fix, which testing techniques that need running or committed code cannot do.
Why A is correct: Threat modeling is the design-phase activity that maps data flows and trust boundaries to expose architectural weaknesses such as missing authorisation, so it identifies and ranks design flaws exactly when the requirement demands, before implementation begins.
Why B is wrong: Dynamic testing is valuable and tempting because it finds real exploitable issues, but it runs against built, deployed code and so cannot surface the design-level trust-boundary flaws the requirement targets before code is written.
Why C is wrong: Static analysis catches insecure code patterns and feels like the earliest control, but it operates on committed source rather than the architecture, so it misses design flaws that exist before code and that no scanner reads from a diagram.
Why D is wrong: A penetration test gives independent assurance and is appealing as a gate, but it happens against a near-complete system late in the lifecycle, so it finds design flaws only after they are expensively built rather than before code is written.