CertSafari
    Snowflake SnowPro Advanced: Security Engineer (SEA-C01)· Lessons

    Domain 4 · Lesson 16/21

    Risk assessment with Horizon Catalog and the Trust Center

    Perform risk assessment and manage risk.

    22 min read
    4.5% of exam
    6 sources
    Published 20 Oct 2026
    Docs as of 4 Oct 2026

    What you will be able to do

    • Use Horizon Catalog capabilities as the governance baseline to measure security risk against
    • Grant the correct Trust Center application role for viewing findings versus managing scanners and violations
    • Choose the scanner package that covers a given risk, and tell violations apart from detections
    • Assess what a partner data share exposes, and verify what each consumer account actually sees
    • Rank findings by likelihood and impact, then remediate, verify and track the mitigation over time
    • Divide risk mitigation between viewer and admin roles so that strategy, implementation and monitoring each have an owner

    Key concept

    Violation versus detection — A Trust Center violation is a standing configuration that fails a scanner's requirement and stays open until the configuration changes. A detection is a single past event that you investigate and learn from, because there is nothing to remediate directly. Everything about ranking and mitigating risk depends on which kind of finding you are looking at.

    1.Horizon Catalog as the governance baseline

    To assess risk, you first need a picture of what good governance looks like, so that you can measure each gap against it. In Snowflake, Horizon Catalog provides that picture. It is the catalog that brings discovery, lineage, protection policies and posture monitoring together, and it promises enterprise-grade governance covering sensitive data protection, data quality, end-to-end lineage, AI guardrails and AI governance. Horizon exists because fragmented security leaves gaps: when access controls and lineage live in separate systems, sensitive data gets hard to track as it moves.

    Each Horizon capability answers one of the questions a risk assessment asks. Classification tells you where sensitive data is. Lineage tells you where that data flows. Masking and row access policies show what protects it. The Trust Center shows whether the configuration has drifted. Role-based access control, Time Travel and access history let you look back at past activity and earlier data states, which you need when a finding has to be confirmed or ruled out.

    Horizon Catalog capabilities mapped to the risk question each one answers
    Risk questionHorizon capabilityWhat it provides
    Where is the sensitive data?Sensitive data classificationAutomatically discovers and classifies columns, then applies tags and policies as schemas evolve
    Where does it flow?End-to-end data lineageColumn-level lineage across Snowflake, external databases, BI tools and OpenLineage feeds
    What protects it?Data protection policiesMasking and row access policies, enforced across Iceberg REST Catalog-compatible engines
    Has the configuration drifted?Access control and Trust CenterContinuous monitoring for misconfigured roles and unprotected columns

    Where a control is enforced decides how much one review can prove. Horizon governance policies run in the query engine, not in the application, so they apply to every caller: a human analyst, a BI tool or an AI agent. AI workloads have no separate governance configuration. One review of a policy therefore covers every client that reaches the data. The same reach extends beyond the account, because shared data products carry their tags and permissions with them.

    Horizon also measures itself. Through Snowflake CoCo, the AI chat assistant in Snowsight that works from natural-language instructions and also guides Trust Center remediation, you can assess governance and observability maturity with scoring skills and receive prioritized recommendations, with no SQL required. Use that score as a starting point for the risk register, not as the register itself.

    Checkpoint 1 of 10· Match them up

    During a risk assessment, each Horizon Catalog capability answers a different question. Match each capability to the risk question it answers.

    Tap a term, then the definition that fits it.

    Sources12

    2.Who can see and who can change Trust Center findings

    Horizon reports posture through the Trust Center, which evaluates each account against the recommendations defined in its scanners and is used to evaluate, monitor and reduce security risk. Access is controlled by two application roles in the SNOWFLAKE database, not by system roles. A user with ACCOUNTADMIN grants SNOWFLAKE.TRUST_CENTER_VIEWER or SNOWFLAKE.TRUST_CENTER_ADMIN to a custom role. In an organization account, the grant comes from GLOBALORGADMIN instead.

    Minimum Trust Center application role per task
    TaskTabMinimum application role
    View account posture and generate a posture reportOverviewSNOWFLAKE.TRUST_CENTER_VIEWER
    View detection findingsDetectionsSNOWFLAKE.TRUST_CENTER_VIEWER
    View violation findingsViolationsSNOWFLAKE.TRUST_CENTER_VIEWER
    Manage violation findings lifecycleViolationsSNOWFLAKE.TRUST_CENTER_ADMIN
    Manage scanner packages and scannersManage scannersSNOWFLAKE.TRUST_CENTER_ADMIN
    View org-level violationsOrganizationORGANIZATION_SECURITY_VIEWER and SNOWFLAKE.TRUST_CENTER_ADMIN

    Because the work divides this way, a security operations team can read everything without being able to change anything. Analysts get a role holding TRUST_CENTER_VIEWER, so they see violations, detections and posture but cannot change scanner schedules or violation status. Only the people who own remediation get TRUST_CENTER_ADMIN. If you grant the admin role and then try to block individual actions, you have granted more than needed and are trying to undo part of the grant. SECURITYADMIN is not how anyone opens the Trust Center.

    Separate view-only and administrator roles for the Trust Centersql
    USE ROLE ACCOUNTADMIN;
    
    CREATE ROLE trust_center_admin_role;
    GRANT APPLICATION ROLE SNOWFLAKE.TRUST_CENTER_ADMIN TO ROLE trust_center_admin_role;
    
    CREATE ROLE trust_center_viewer_role;
    GRANT APPLICATION ROLE SNOWFLAKE.TRUST_CENTER_VIEWER TO ROLE trust_center_viewer_role;
    
    GRANT ROLE trust_center_admin_role TO USER example_admin_user;
    
    GRANT ROLE trust_center_viewer_role TO USER example_nonadmin_user;

    A mitigation strategy works best when each stage has an owner, and the two application roles map onto the stages. Developing the strategy is viewer work: analysts read violations and detections, rank them, and decide per finding whether to change a configuration, investigate an event, or record a justified resolution. Implementing it is admin work, because only TRUST_CENTER_ADMIN holders can run remediation, manage scanners and change a violation's status. Monitoring never stops. Configure Trust Center notifications so new findings reach the team, then review the account posture over a chosen period to see whether open findings are falling. Splitting the loop this way means the people judging progress cannot quietly close findings themselves.

    Checkpoint 2 of 10· Check yourself

    A security operations team must review Trust Center violations and detections. The team must not be able to resolve violations or change scanner schedules. What is the least-privilege way to give them access?

    Checkpoint 3 of 10· Exam question

    A security operations group needs analysts to see Trust Center violations, detections and account posture, but analysts must not change scanner schedules or violation status. What is the correct grant?

    Sources3

    3.Scanner packages and what each one looks for

    A scanner is a background process that checks the account for two things: how it is configured, and anomalous events. Scanners are grouped into packages. You must enable a package before you can change the schedule of any scanner in it. Scanners also run in one of two ways. Schedule-based scanners run at set times. Event-driven scanners respond to events, such as a sign-in from an unusual IP address, so they cannot be scheduled, and the Trust Center reports their detections within an hour of the event.

    The CIS Benchmarks package checks configuration against the CIS recommendations. Read its results literally. For recommendation 4.10, the account gets a violation if it has no masking policy at all, but the scanner does not evaluate whether sensitive data is protected appropriately or whether the policy is assigned to any table. Recommendation 4.11 checks row access policies in the same existence-only way. A clean CIS result therefore lowers the likelihood of some misconfigurations. It does not prove that sensitive columns are protected.

    The Threat Intelligence package targets identity risk. It flags risk based on four criteria:

    • User types: whether a user is a human or a service.
    • Authentication methods or policies: whether a user signs in with a password without being enrolled in MFA.
    • Login activity: whether a user has not signed in recently.
    • Abnormal failure rates: whether a user has a high number of authentication failures or job errors.

    Its individual scanners turn those criteria into specific findings. One identifies legacy service users who have a password and have signed in with only a password. Another identifies users with a high volume of authentication failures, which might indicate attempted account takeovers, misconfigurations, exceeded quotas or permission issues. An event-driven scanner flags sign-ins from users who had not signed in for 90 days. So if the goal is to find password-only service users and spikes in failed logins, enable Threat Intelligence. The AI Security package covers something else: Cortex Agents, AI scanner coverage and Cortex AI Guardrails.

    Checkpoint 4 of 10· Check yourself

    Your company has started building Cortex Agents. The security lead wants Trust Center findings on agent misconfigurations and wants to see whether Cortex AI Guardrails are switched on. Which scanner package should be enabled?

    Checkpoint 5 of 10· Exam question

    A security team must find service accounts that still authenticate with passwords and detect abnormal spikes in failed logins across the account. Which Trust Center scanner package should they enable?

    Sources32

    4.Violations and detections behave differently

    A violation describes the account as it is now. A scanner can look at an entity at any time and decide from its current configuration alone whether it is in violation. It keeps reporting the violation until someone changes that configuration. Users without MFA are the standard example: the finding stays open for as long as those users remain.

    A detection describes something that has already happened: a sign-in from an unrecognized IP address, a large transfer of data to an external stage, a task with a high error rate. Each event produces its own detection. Because the event is in the past, there is nothing to remediate directly. You investigate whether it matters, and if it does, you act to prevent similar events in the future.

    How the two finding types differ
    PropertyViolationDetection
    What it representsA configuration that persists over timeA one-time, unique event
    RemediationChange the configuration; the next scan clears itNot possible directly; investigate and prevent recurrence
    Lifecycle managementTriage, resolve or reopen with justificationCannot be muted, resolved or reopened
    Organization tabAggregated across accountsNot viewable at the organization level
    SeverityLow, medium, high or criticalFilterable by Critical, High, Medium or Low

    Two scope limits matter during an assessment. Reader accounts are not supported, so a reader account you create for a partner gets no Trust Center findings of its own. The Organization tab can show violations per account, but it cannot resolve or reopen them. You have to sign in to the account that has the violation to do that.

    Checkpoint 6 of 10· Check yourself

    The Detections tab shows a High-severity finding: a sign-in from an unrecognized IP address yesterday. The security lead asks what the team should do with this finding. Which response fits how the Trust Center handles it?

    Checkpoint 7 of 10· Exam question

    Which TWO statements about Trust Center findings are accurate?(Select 2)

    0 of 2 selected

    Sources3

    5.Assessing a partner data sharing agreement

    A sharing agreement creates exposure that grows without any new approval. With shares, listings and reader accounts, the provider decides which accounts consume a share by adding them to it. New objects added to a share become available to every consumer immediately. So assess a share on two things: what it contains today, and what someone could add to it tomorrow. Review three things for each share:

    • Which objects are granted, directly or through a database role, and whether each one is a secure object.
    • Which consumer accounts are added, and whether any of them are reader accounts, which can consume data only from the provider that created them.
    • Whether any policy-protected data in the share filters or masks per consumer, and how that was tested.

    Shared objects are read-only for consumers, and the provider can revoke access to a share at any time. That caps the damage a partner can do to your data, and it gives you a fast way to contain a problem. It does nothing to limit what a partner can read while access is granted. That limit comes from the policies.

    Per-partner protection has to be built from information that still means something in the consumer's account. Dynamic Data Masking policies and row access policies can call IS_DATABASE_ROLE_IN_SESSION to check for a shared database role. The provider creates that database role in the same database as the protected table and grants it to the share. When the consumer creates a database from the share, the database role goes to the role that created it, and the policy condition can then be met. A partner holding the role sees what the policy allows. A partner without it sees the masked or filtered result.

    Never assume a per-account filter works. Test it. The SIMULATED_DATA_SHARING_CONSUMER session parameter lets you, in the provider account, query a secure view as if you were a user in a given consumer account. It supports only secure views and secure materialized views, not secure UDFs. Repeat the simulation once for each partner account on the share, and treat any row a partner should not see as an open finding.

    Simulate querying shared secure views as consumer account xy12345sql
    ALTER SESSION SET SIMULATED_DATA_SHARING_CONSUMER = xy12345;

    Checkpoint 8 of 10· Check yourself

    Last quarter, your team approved a share that gives a partner one secure view. A colleague now says the share is low risk for good: partners can't modify shared objects, and the approval covered exactly what the partner sees. What is wrong with that reasoning?

    Sources456

    6.Ranking vulnerabilities by likelihood and impact

    A list of findings is not yet a ranked list of risks. To rank one, ask how likely it is to be exploited and how much would be lost if it were, then work on the findings where both are high first. The Trust Center scanner packages, findings and CIS Benchmarks supply the evidence for both questions. The severity label is the scanner's own rating and a reasonable first sort. Ranking means refining that sort with what you know about the account.

    Trust Center evidence and how it informs likelihood or impact
    EvidenceBearing on likelihoodBearing on impact
    Open violationExposure stays open until the configuration changesDepends on what the misconfigured entity can reach
    DetectionThe event already occurred; investigate whether it is meaningfulShown by what the event touched, such as data moved to an external stage
    High volume of authentication failuresMight indicate attempted account takeoverDepends on the user's grants and user type
    At-risk entities by type (User, Table, Procedure)More entities means more paths to exploitWider spread of affected objects
    Data in a shareEvery consumer account added to the share can read itAll consumers get new objects immediately

    Here is a worked example. Two findings arrive in the same week. The first is a CIS violation because the account has no row access policy. The second is a Threat Intelligence finding: a legacy service user that still signs in with only a password, plus a high volume of authentication failures on that same user. The CIS item is an existence check, and lineage and classification may show that the sensitive tables are already protected by masking, which lowers its impact. The service user combines a weak authentication method with active failed attempts, so its likelihood is high. Its impact is whatever that user's grants can reach. It goes to the top of the list.

    Ranking also tells you which kind of mitigation to use. A violation needs a configuration change that a scanner can confirm. A detection needs investigation, and if it turns out to matter, a preventive control so that similar events stop happening. Sort the Violations tab by severity, scanner package, status or scan time, then record your own ranking as triage notes on each violation. The reasoning then stays attached to the finding.

    Checkpoint 9 of 10· Check yourself

    After ranking, the item at the top of your list is an open violation for users who sign in with a password and no MFA. Management wants evidence that this risk has been mitigated. Which plan provides that evidence?

    Sources32

    7.Implementing, verifying and monitoring mitigations

    Mitigation starts on the Remediation tab of a violation, which lists the steps to fix it. Some violations support one-click remediation. The Trust Center builds a plan that shows the potential impact, the affected users and the SQL it will run in order, and you approve it with Execute plan. Review that SQL first. If a statement fails, the plan stops there. Statements that already succeeded stay applied, and the result is reported as Violation partially remediated. Violations without one-click remediation can go to CoCo. That requires TRUST_CENTER_ADMIN plus the SNOWFLAKE.CORTEX_USER database role. Some fixes can't be made in SQL at all, such as an organization-wide change to MFA policy.

    Confirm every fix with a scan. Run the scanner on demand instead of waiting for its next scheduled run. A remediated finding can still appear open in Snowsight for up to 3 hours after the rerun, so a finding that stays visible right after the fix has not necessarily failed.

    Track decisions along with fixes. On the Violations tab you can triage a violation, record evidence or progress notes, and resolve or reopen it with a justification for auditors. Resolving a violation suppresses its email notifications while the work continues. It does not change the configuration, so the risk remains until a scan confirms the fix.

    Monitoring over a period uses the Account posture section on the Overview tab. It shows scanning status, open findings and at-risk entities broken out by type, and it produces a PDF report. In the report, violations are grouped as newly opened, remediated, increased, decreased, unchanged or muted, and scanners that ran clean are listed alongside open findings. Together these show auditors that your mitigation is making progress, not just how many findings are open on one day. The report's configuration-change section is limited to package enablement and triage activity. Other changes are in query history.

    Checkpoint 10 of 10· Put it in order

    A violation on the Violations tab supports one-click remediation, and you hold TRUST_CENTER_ADMIN. Arrange the steps so the fix is applied safely and you can prove the risk is gone.

    1. 1.Select the violation and open its remediation, so the Trust Center generates a remediation plan
    2. 2.Run the scanner on demand, then confirm the violation no longer appears (allow up to 3 hours for Snowsight to update)
    3. 3.Review the plan's potential impact, affected users and SQL statements, in the order they will run
    4. 4.Select Execute plan and check whether the result says remediated, partially remediated or failed

    Sources23

    Practise it for real

    Set up read-only risk review and verify what one partner account sees through a shared secure view.

    1. 1.As ACCOUNTADMIN, create trust_center_viewer_role, grant it the SNOWFLAKE.TRUST_CENTER_VIEWER application role, and grant the role to an analyst user.

      Why: Analysts can read findings without being able to change scanner schedules or violation status.

      You should see: The analyst can open the Overview, Violations and Detections tabs but cannot manage scanners or change a violation's status.

    2. 2.With a TRUST_CENTER_ADMIN role, enable the Threat Intelligence scanner package on the Manage scanners tab.

      Why: This package flags password-only service users and abnormal authentication failure rates.

      You should see: Schedule-based scanners run on their schedules. Event-driven scanners report detections within an hour of an event.

    3. 3.In the provider account, run ALTER SESSION SET SIMULATED_DATA_SHARING_CONSUMER = xy12345, replacing xy12345 with the partner's account name, and query a secure view that is in the share.

      Why: This tests the per-partner filter before you rely on it.

      You should see: Only the rows that partner account should see are returned. Any other row is a finding to record and fix.

    Sources

    (6)

    Every claim above is drawn from one of these pages, quoted as it was written on the date shown.

    1. 1.
      “trust every answer with enterprise-grade governance: sensitive data protection, data quality, end-to-end lineage, AI guardrails, and AI governance.”
      ↩︎ Horizon Catalog as the governance baseline
      “Governance policies execute at the query engine layer, not the application layer.”
      ↩︎ Horizon Catalog as the governance baseline
      “Trust Center continuously monitors for misconfigured roles and unprotected columns.”
      ↩︎ Horizon Catalog as the governance baseline
      “Shared data products carry their tags and permissions.”
      ↩︎ Horizon Catalog as the governance baseline
      “assess your governance and observability maturity with scoring skills, and get prioritized recommendations”
      ↩︎ Horizon Catalog as the governance baseline
      “Configure governance policies in natural language”
      ↩︎ Horizon Catalog as the governance baseline
    2. 2.
      “CoCo opens a chat that explains the violation finding in the context of your account, recommends remediation steps”
      ↩︎ Horizon Catalog as the governance baseline
      “It surfaces the list of Cortex AI agents, AI Security scanner coverage, enablement status of Cortex AI guardrails”
      ↩︎ Scanner packages and what each one looks for
      “broken out by entity type such as User, Table, and Procedure”
      ↩︎ Ranking vulnerabilities by likelihood and impact
      “Statements that already ran successfully stay applied, and the result is reported as Violation partially remediated.”
      ↩︎ Implementing, verifying and monitoring mitigations
      “After completing a remediation, you can verify the fix by running the scanner on demand rather than waiting for the next scheduled run.”
      ↩︎ Implementing, verifying and monitoring mitigations
      “might appear as an open finding in Snowsight for up to 3 hours”
      ↩︎ Implementing, verifying and monitoring mitigations
      “Scanners that ran clean during the period are reported alongside open findings, so a reviewer can see which checks your account satisfies.”
      ↩︎ Implementing, verifying and monitoring mitigations
      “After completing a remediation, you can verify the fix by running the scanner on demand rather than waiting for the next scheduled run.”
      ↩︎ Checkpoint
      “Always review the SQL statements in the plan before you select Execute plan.”
      ↩︎ Checkpoint
    3. 3.
      “a user with the ACCOUNTADMIN role must grant the SNOWFLAKE.TRUST_CENTER_VIEWER or SNOWFLAKE.TRUST_CENTER_ADMIN application role to your role.”
      ↩︎ Who can see and who can change Trust Center findings
      “use the GLOBALORGADMIN role, not ACCOUNTADMIN, to grant the Trust Center application roles.”
      ↩︎ Who can see and who can change Trust Center findings
      “You can also use the Trust Center to configure proactive notifications that help you monitor your account for security risks.”
      ↩︎ Who can see and who can change Trust Center findings
      “The Trust Center reports detections that are generated by event-driven scanners within an hour of the time that an event occurs.”
      ↩︎ Scanner packages and what each one looks for
      “You must enable a scanner package before you can change the schedule for a scanner.”
      ↩︎ Scanner packages and what each one looks for
      “Authentication methods or policies: Whether a user logs in to their account with a password without being enrolled in MFA.”
      ↩︎ Scanner packages and what each one looks for
      “Identifies legacy service users who have a password and (a) have signed in with only a password”
      ↩︎ Scanner packages and what each one looks for
      “which might indicate attempted takeovers of an account, misconfigurations, exceeded quotas, or permission issues”
      ↩︎ Scanner packages and what each one looks for
      “Scanners continue to report on violations unless you change the configuration to remediate the violations.”
      ↩︎ Violations and detections behave differently
      “A large amount of data was transferred to an external stage.”
      ↩︎ Violations and detections behave differently
      “Snowflake reader accounts aren’t supported.”
      ↩︎ Violations and detections behave differently
      “Currently, you can’t view detection findings at the organization level.”
      ↩︎ Violations and detections behave differently
      “A graph of scanner violations over time, color-coded by low, medium, high, and critical severity.”
      ↩︎ Ranking vulnerabilities by likelihood and impact
      “Sort or filter violations by severity, scanner package, scanner version, scanned time, updated time, or status.”
      ↩︎ Ranking vulnerabilities by likelihood and impact
      “Based on the information that the Trust Center provides, you can investigate whether the detection is meaningful.”
      ↩︎ Ranking vulnerabilities by likelihood and impact
      “Resolve or reopen violations for any reason and record justification for audit needs.”
      ↩︎ Implementing, verifying and monitoring mitigations
      “A detection occurs one time and represents a unique event.”
      ↩︎ Key concept
      “Currently, you can’t manage the lifecycle of a detection.”
      ↩︎ Exam trap 1
      “Email notifications are suppressed for resolved violations.”
      ↩︎ Exam trap 2
      “does not evaluate whether sensitive data is protected appropriately”
      ↩︎ Exam trap 3
      “a user with the ACCOUNTADMIN role must grant the SNOWFLAKE.TRUST_CENTER_VIEWER or SNOWFLAKE.TRUST_CENTER_ADMIN application role to your role.”
      ↩︎ Exam trap 4
      “View violation findings | Violations | SNOWFLAKE.TRUST_CENTER_VIEWER | SNOWFLAKE.TRUST_CENTER_ADMIN role can also view violations.”
      ↩︎ Checkpoint
      “After you remediate a violation, the violation still appears in the Violations tab until the next scheduled run of the scanner package”
      ↩︎ Prediction
      “If the detection is meaningful, you can take actions to prevent similar events in the future.”
      ↩︎ Checkpoint
    4. 4.
      “New objects added to a share become immediately available to all consumers, providing real-time access to imported data.”
      ↩︎ Assessing a partner data sharing agreement
      “Access to a share (or any of the objects in a share) can be revoked at any time.”
      ↩︎ Assessing a partner data sharing agreement
      “a reader account can only consume data from the provider account that created it.”
      ↩︎ Assessing a partner data sharing agreement
    5. 5.
      “The provider defines the policy to call the IS_DATABASE_ROLE_IN_SESSION function to evaluate the shared database role”
      ↩︎ Assessing a partner data sharing agreement
      “Grant the database role to the share containing the protected table.”
      ↩︎ Assessing a partner data sharing agreement
    6. 6.
      “The SIMULATED_DATA_SHARING_CONSUMER session parameter only supports secure views and secure materialized views, but does not support secure UDFs.”
      ↩︎ Assessing a partner data sharing agreement
      “Do not include secure objects that use the CURRENT_USER or CURRENT_ROLE functions in their definition.”
      ↩︎ Prediction

    Ready to test yourself?

    Practise the 16 questions on this subdomain.

    Spotted a mistake, or was something unclear? Tell us.