20 free SC-900 practice questions with answers and explanations, covering security and identity concepts, Microsoft Entra, Microsoft security solutions and Microsoft Purview compliance. A good first exam for anyone heading into security.
Study the full SC-900 course, free to start.
Official SC-900 study guide on Microsoft Learn
Question 1. Fabrikam is rolling out an AI assistant that can read documents from its SharePoint sites. The project team assumes security is handled for them because the provider hosts the model. Which task still belongs to Fabrikam?
Answer: D, Deciding which staff may use the assistant, which data sources it may reach, and how prompt injection is monitored. Under the AI shared responsibility model the provider secures the AI platform: infrastructure, model hosting and platform-level safety controls. The customer keeps data protection, identity and access management, safe configuration of connectors and data sources, mitigation of AI-specific risks such as prompt injection, and user training and acceptable-use policies.
Question 2. An administrator no longer holds a standing privileged role. Instead they request elevation when a task requires it, receive only the permissions that task needs, and lose the role again afterwards. Which Zero Trust guiding principle is this?
Answer: C, Use least privileged access. Least privileged access grants only the minimum access needed, for only as long as it is needed, implemented through just-in-time (JIT) and just-enough-access (JEA) and reinforced by risk-based adaptive policies. Verify explicitly is about authenticating and authorising on all available signals. Assume breach is about segmentation, end-to-end encryption and monitoring. Trust but verify is not a Zero Trust principle and is close to the opposite of one.
Question 3. Two organisations need to exchange a large encrypted file every night. They have no secure channel to agree a shared secret in advance, and the file is far too big to be worth encrypting with a slow algorithm. What is the usual way to satisfy both constraints?
Answer: D, Use asymmetric encryption to exchange a symmetric key safely, then encrypt the file itself with that symmetric key. Symmetric encryption uses one key for both directions and is computationally fast, which is what large volumes need, but both parties must hold the same secret and distributing it is the hard part. Asymmetric encryption uses a mathematically related public and private key pair, so the public key can be shared freely and the distribution problem goes away. Combining the two, which is what HTTPS does, gives the speed of one and the safe key exchange of the other. Emailing a key puts the secret in the least trustworthy place available, and a hash cannot be reversed at all, salted or not.
Question 4. After an incident, an organisation cannot establish which accounts opened a sensitive file share, or from where. Sign-ins were correctly verified and permissions were correctly assigned. Which pillar of its identity infrastructure is weak?
Answer: D, Auditing, because there is no record of who did what, when and from where. The four pillars of an identity infrastructure are administration, authentication, authorisation and auditing. Auditing covers sign-in logs, activity reports, alerts on anomalous behaviour and forensic investigation, and it is what an organisation depends on to establish the scope of a breach and to meet compliance obligations. The scenario states that authentication and authorisation both worked, which leaves the pillar teams most often neglect.
Question 5. Two engineering firms are collaborating and need each other's staff to reach a shared design portal. The IT team wants to avoid creating and managing hundreds of individual accounts for partner employees. Which approach fits?
Answer: B, Establish federation, configuring a trust relationship between the two organisations' identity providers so partner staff sign in with their own credentials. Federation enables access across organisational boundaries by establishing trust between separate identity providers, so users authenticate with credentials they already hold and no new accounts are needed. Duplicate accounts recreate exactly the administrative burden the team wants to avoid, a shared account destroys accountability and makes auditing meaningless, and an unlisted public link is not access control at all. Note that the trust configured here works in one direction unless both organisations set it up.
Question 6. An AI agent runs on a schedule overnight to reconcile invoices. No person is signed in when it runs, and it uses its own assigned roles and permissions to reach the finance system. Which authentication pattern is this?
Answer: C, Unattended, because the agent acts under its own authority with no user present. Unattended is where the agent acts as an autonomous identity using its own roles and permissions, with no user involved. Attended, also called on-behalf-of, is where the agent acts for a signed-in human using delegated permissions that user granted, so the agent can only reach what the user could. Having a sponsor does not make an agent attended: the sponsor is the human accountable for the agent, which is a governance relationship rather than an authentication one.
Question 7. A new starter arrives on their first day with no registered authentication methods, and the organisation wants them to go straight to a passwordless credential without ever being issued a password. Which method solves the bootstrapping problem?
Answer: C, A Temporary Access Pass issued by an administrator. A Temporary Access Pass is a time-limited passcode issued by an administrator that can be used to sign in and register other authentication methods, including passwordless ones, which is exactly the onboarding case. It also covers a user who has lost their credentials. Email one-time passcode is a secondary method for self-service password reset and guest sign-in. SMS sign-in would still need registering first. Security questions cannot start a sign-in and are being retired for SSPR in March 2027.
Question 8. An organisation with cloud-only user accounts on Microsoft Entra ID Free wants to block passwords containing its own brand names. What does it need?
Answer: B, Microsoft Entra ID P1 or P2, because the custom banned password list requires a premium licence. The global banned password list is applied automatically to every user in every tenant at no cost, cannot be disabled, and is built from Microsoft's own telemetry on real password spray attacks, so it holds common weak passwords rather than one organisation's brand names. The custom banned password list is the feature for organisation-specific terms, and for cloud-only users it requires Microsoft Entra ID P1 or P2. The on-premises components extend the same lists to Active Directory, they do not create them.
Question 9. An organisation wants an administrator who can manage one named expense reporting application and nothing else. No built-in role is that narrow. What does implementing this involve?
Answer: B, Creating a custom role definition from the preset permission list, then creating a role assignment at object scope for that application. Granting permission with a custom Microsoft Entra role is a two-step process: first create the role definition as a collection of permissions chosen from the same preset list the built-in roles use, then create a role assignment giving that definition to a user or group at a chosen scope. Scope can be organisation-wide or an object scope such as a single application. Built-in roles have a fixed permission set that cannot be modified, and Azure RBAC governs Azure resources rather than Entra applications. Custom roles require Entra ID P1 or P2.
Question 10. An analyst notices that thousands of failed sign-ins against a user account produced no Microsoft Entra ID Protection risk detections at all. Why?
Answer: C, ID Protection only generates risk detections when the correct credentials are used in the authentication request. Identity Protection only generates risk detections when correct credentials are used, because until someone is actually authenticating with a working credential there is no risk of credential compromise to report. Failed attempts with wrong credentials still appear in sign-in logs and still matter operationally, they simply are not what ID Protection scores. ID Protection covers both user identities and workload identities, and detections run in real time at each sign-in.
Question 11. A media company serves the same web application from several regions and wants attacks blocked as close to the source as possible, before the traffic reaches any of its regional infrastructure. Which Azure WAF deployment option fits?
Answer: C, WAF on Azure Front Door, because it protects at the network edge across all regions. Azure Front Door is Microsoft's global content delivery and application delivery platform, and WAF on Front Door provides global protection by blocking attacks closer to the source before traffic reaches your infrastructure. Application Gateway provides regional protection for applications behind the gateway, which suits a single-region deployment. Application Gateway for Containers is about containerised workloads specifically, and DDoS Protection covers layers 3 and 4 rather than web exploits.
Question 12. A retailer runs Azure SQL, an open-source relational database and Azure Cosmos DB, and wants attack detection and threat response across all three. Which Microsoft Defender plan should it enable?
Answer: A, Microsoft Defender for Databases, which protects the database estate including Azure SQL, open-source relational databases and Azure Cosmos DB. Cloud workload protection is delivered through Microsoft Defender plans specific to the resource types in your subscriptions, and Defender for Databases is the plan that protects the database estate with attack detection and threat response for Azure SQL, open-source relational databases and Azure Cosmos DB. Defender for Storage covers Azure Storage accounts, Defender for Servers covers Windows and Linux machines, and Defender for Resource Manager monitors management operations.
Question 13. Over six weeks an attacker performs quiet reconnaissance, then establishes persistence, then slowly moves data out. Each individual event is low severity and no single analytics rule fires convincingly. Which Microsoft Sentinel capability is designed to surface this?
Answer: D, Fusion, the correlation engine that links low-severity signals from multiple sources into high-confidence incidents. Fusion is the correlation engine in the AI and machine learning detection approach. It automatically links low-severity signals from multiple sources into high-confidence, actionable incidents, which is how advanced multistage attacks get surfaced when no individual rule would catch them. Threat intelligence matching only helps if the indicators are already known. Content hub is a distribution mechanism for prebuilt content, and automation rules manage how incidents are handled after they exist.
Question 14. An analyst needs to see the full scope of a breach on a laptop in near real time, prioritise the alerts, and take response actions to remediate the device. Which Microsoft Defender XDR capability provides this?
Answer: C, Endpoint detection and response in Microsoft Defender for Endpoint, which provides near real-time actionable attack detections. Endpoint detection and response provides advanced attack detections that are near real time and actionable, letting security analysts prioritise alerts, see the full scope of a breach and take response actions to remediate threats. Attack surface reduction is the preventative first layer, next generation protection is the antivirus layer, and Microsoft Secure Score for Devices is a posture measure. Endpoint detection and response is a Defender for Endpoint Plan 2 capability.
Question 15. A third-party productivity app was granted broad OAuth permissions to read mail and files on users' behalf two years ago, and nobody has reviewed it since. Which Microsoft Defender for Cloud Apps capability is meant to manage this?
Answer: B, App governance, which gives visibility and control over OAuth apps and monitors current and expired credentials. OAuth apps often hold extensive permissions to access data in other apps on behalf of a user, which makes them attractive to attackers. App governance closes that gap, giving in-depth visibility and control over OAuth apps integrated with Microsoft Entra ID, Google and Salesforce, watching for unused apps and monitoring both current and expired credentials. Cloud discovery finds shadow IT, SSPM addresses configuration, and information protection addresses the data itself.
Question 16. A new analyst can browse parts of the Service Trust Portal from a personal browser session, but a colleague reports that several documents are not visible to him at all. What is the most likely explanation?
Answer: C, Some resources require signing in as an authenticated user with a Microsoft Entra organisation account, and the Resources for your Organization category is further restricted by tenant subscription and permissions. Access to some Service Trust Portal resources requires signing in with a Microsoft cloud services account, meaning a Microsoft Entra organisation account, and entra accounts associated with organisations reach the full range of documents. On top of that, the Resources for your Organization section lists documents restricted by tenant, based on subscription and permissions. Age of the document, an E5 requirement and a professional services agreement are none of them the gate here.
Question 17. A hospital wants a match on a patient record number to mean a real patient in its own records, not merely any number of the right shape. Which category of sensitive information type should it use?
Answer: D, An exact data match type, shown in the Purview portal as an EDM classifier, referencing the hospital's own database of values. Exact data match based classification creates custom types that reference exact values from a database of sensitive information, which is what makes it right for highly specific data such as patient record numbers, employee IDs or proprietary product codes. A built-in or custom pattern-based type matches the shape of a number and cannot tell a real record from a coincidence, and named entity types detect person names, physical addresses and medical terms and conditions rather than record identifiers.
Question 18. A security team wants protective controls that tighten automatically for a user whose behaviour has become risky and relax again once it settles, without an administrator rewriting policies each time. What provides this?
Answer: C, Adaptive protection, which uses insider risk levels to drive controls in DLP, Data Lifecycle Management and Conditional Access. Adaptive protection is the feature that makes protection dynamic. Insider risk management calculates elevated, moderate or minor risk levels from behaviour, and adaptive protection applies matching controls through Purview DLP, Purview Data Lifecycle Management and Microsoft Entra Conditional Access. Levels update continuously and the controls follow, with no manual intervention. Mandatory labelling, retention policies and sensitive information types are all static configurations that do not respond to a change in a user's risk.
Question 19. Every document on a project site must be kept for five years, but a subset of contracts stored on that same site must be kept for ten. What is the least effort way to configure this?
Answer: B, Apply a five-year retention policy to the site, then apply a ten-year retention label to the individual contracts. Retention policies assign the same settings at site or mailbox level and items inherit them from the container, so they are the efficient way to cover everything. Retention labels assign settings at item level, so they are the way to give particular documents a longer period. This is the exact scenario the documentation uses. Stacking labels is not possible, because an email or document can carry only one retention label at a time.
Question 20. A security officer is worried that granting somebody a role in Microsoft Purview Data Map would expose confidential customer records. What is the correct reassurance?
Answer: A, Everything held in Data Map is metadata describing the data, and none of its permissions or roles provide access to the underlying data itself. All data in Data Map is metadata, descriptive information about your data rather than the data itself, and none of the permissions or roles in Data Map provide access to the actual data. Scanning captures structure, location and relationships, and classification records what kinds of sensitive information an asset contains, not the values. Data Map does not store copies of the source files and it is not limited to sources marked public.
It's a fundamentals exam that asks you to describe what each Microsoft security, compliance and identity product does. The challenge is breadth rather than depth: there are a lot of products with similar names to keep straight.
Most people studying part-time plan two to four weeks at 20 to 30 minutes a day, with the most time on Microsoft security solutions, the largest area.
No. Microsoft exam questions are confidential, and sharing them breaks the agreement every candidate accepts. These are original questions written to the published skills outline, so they test the same knowledge in the same style.
You can enrol free with no card. Free accounts get the opening modules of every course; Pro (£9.99 a month, or £79.99 a year) unlocks every module, the full timed mock exam and an adaptive study plan.
Answer each one before you look at the explanation, and keep a list of the ones you get wrong. That list is your study plan: it's made of exactly the things you don't know yet.
CertBuddi is an independent study aid, not affiliated with or endorsed by Microsoft. These are original practice questions, not real exam questions.