GARP ICBRR Exam Actual Questions
International Certificate in Banking Risk and Regulation (ICBRR) (Page 19 )

• Why: If an application on EC2 needs to call other AWS services, using an IAM role (attached to the EC2 instance as an instance profile) provides temporary, automatically rotated credentials. This avoids embedding access keys in the application.
• How it works: Create a role with the minimum required permissions, attach it to the EC2 instance, and let the AWS SDK used by the application retrieve credentials from the instance metadata service. The credentials rotate automatically and are scoped to the policy you define.
• Why not other options:

• runParallel is defined as async, so it returns a Promise.
• To use the result, you attach a then handler (or use await inside another async function).

• A: “Async runParallel() .then(data);” is invalid. “async” is a keyword for function declarations, not something you put in front of a call, and you’d need a function as the then-argument.
• B: “runParallel().done(function(data){return data;});” .done is not standard in native Promises (it exists in some libraries). Without a compatible library, this won’t work.
• C: “runParallel().then(data);” is invalid because then expects a function as its argument; data is not a function.
• D: “runParallel().then(function(data){ return data; });” is valid (or equivalently, runParallel().then(data => data);).

• The robust, standard option is D.
• B could work only with a library that provides .done; otherwise it’s not reliable.

• The two correct options are:

• The two incorrect options:

• A recipient who already has a Microsoft account or an organizational account can authenticate by signing in.
• A recipient without a usable Microsoft identity may receive a one-time passcode by email.
• The passcode is sent to the same email address that was invited and is used to verify ownership of that address.

• IPv6 address types you’ll typically see:

• Unique Local Addresses (ULAs)

• If an address begins with fdXX:… or fc00:…, place it under “Unique Local Address” (ULA).
• If it begins with 2000::/3, place it under “Global Unicast.”
• If it begins with fe80::, place it under “Link-local.”
• If it begins with ff00::, place it under “Multicast.”

• SageMaker endpoint: Not required because the model is being used through Amazon Bedrock.
• SageMaker Model Registry: Used for managing SageMaker models, not for making a Bedrock custom model available.
• Grant access in Amazon Bedrock: IAM permissions may still be needed, but granting access alone does not provide the required inference capacity.

• Post-Incident Review (A) is the after-action step used to capture lessons learned, identify process gaps, and drive improvements across the incident response lifecycle. This directly supports process maturity and a stronger security posture.
• Incident Analysis (D) focuses on investigating root causes during/after an incident, which is important but more of a diagnostic activity than a standalone baseline improvement to maturity.
• Fast Eradication (B) and Incident Containment (C) are tactical actions aimed at stopping the incident, not on maturing the process.

• Basic: Self-help only; no technical support or design assessment.
• Standard: Technical support, but not the full proactive architectural guidance required here.
• Professional Direct: Provides proactive guidance and architecture-related assistance.
• Premier: Also supports this, but is a more expensive, enterprise-level option.

• AWS IoT Greengrass supports local ML inference using models trained in Amazon SageMaker.
• The model and inference code run on the factory’s local server, so processing continues even when internet connectivity is unavailable.
• A Greengrass component can:

• This provides low latency and avoids sending images to AWS for every inference.

• A: Uses AWS-hosted services and therefore depends on internet connectivity. It would fail during an outage.
• C: Snowball Edge is unnecessary for this use case and adds deployment complexity. A local Greengrass server is the intended edge-inference solution.
• D: Amazon Monitron is designed for monitoring industrial equipment health, not analyzing camera images with a custom SageMaker model.

• Correct answer: D) Amazon Aurora Serverless

• Aurora Serverless is serverless, so compute capacity scales automatically and can pause during idle periods. This minimizes ongoing administration and you pay only for actual usage, which is ideal for infrequent workloads.
• PostgreSQL on EC2 requires full OS and DB administration.
• RDS for PostgreSQL is managed but still involves provisioning capacity and some maintenance tasks.
• Aurora PostgreSQL (non-Serverless) is managed but requires you to manage capacity, though with some automation; it’s not as hands-off as Serverless for intermittent usage.

• A: True. You can change the size of a Virtual Warehouse after creation (via UI or ALTER WAREHOUSE).

• B: True. A warehouse can be resized while it’s running; the resize will take effect without needing to stop queries.

• C: True. A warehouse can be configured to suspend after a period of inactivity (AUTO_SUSPEND).

• D: True. A warehouse can be configured to auto-resume when new queries are submitted (AUTO_RESUME).

• On-Demand: You pay by the hour with no long-term commitment. This is ideal for workloads that are infrequent or have highly variable usage. Since this job runs only 24 hours once per year, you pay only for those 24 hours, with no wasted upfront or long-term payments.
• Reserved Instances: Require a 1- or 3-year commitment and pay for capacity whether you use it or not. For a workload that runs just 24 hours a year, you’d be paying for most of the term unused, making it far less cost-efficient.
• Spot Instances: Can be interrupted whenever AWS needs capacity, which is unacceptable for an uninterruptible workload.
• Dedicated Instances: Generally more expensive and aimed at licensing or compliance requirements; not needed for this use case.

Why

cmd 
PowerShell.exe -ExecutionPolicy Bypass -File C:\Scripts\ConfigureVM.ps1

Why the other options are less suitable

• GPO logon script: Runs only when a user logs on and may not run during initial VM provisioning.
• GPO startup script: Can work, but requires the VM to be joined to the domain and the policy to be applied. It is also normally evaluated at each startup.
• Place scripts in a VHD: Merely placing scripts on a disk does not cause them to execute.

• What D means: In Security Command Center (SCC), a posture combines both preventative controls (predefined and custom organization policies) and detective controls (predefined and custom Security Health Analytics [SHA] modules). Scope this posture to the business unit folder to enforce those controls across the Vertex AI environment.

• Why this is correct: You need both preventative guardrails (policies) and detective guardrails (SHA findings) to meet internal security requirements. A posture provides a unified, scalable way to apply and monitor these controls to a specific organizational unit.

• Why the other options are weaker:

• If you want, I can break down the exact components of a SCC posture and how to apply it to a Vertex AI project.

• Self-signed certificate: Causes a certificate warning, not incorrect IP resolution.
• Nameserver record: A DNS record problem would typically affect multiple users, not just one.
• IP helper: Relays DHCP or other broadcast traffic between networks; it does not control website name resolution.

1. Ingestion Lambda receives requests from API Gateway and places messages in an SQS queue.
2. Processing Lambda is triggered by the queue and loads records into Aurora PostgreSQL.

• Decoupling: API request handling is separated from database loading.
• Buffering: SQS absorbs bursts when incoming traffic exceeds the database’s processing capacity.
• Controlled concurrency: Configure the SQS-triggered Lambda’s concurrency and batch size to avoid overwhelming Aurora or exhausting database connections.
• Automatic retries: Messages that fail processing can return to the queue and be retried. A dead-letter queue can capture persistent failures.
• Lower operational effort: The solution remains serverless and avoids managing EC2 servers or Kubernetes clusters.

• A: EC2 and Tomcat add server-management overhead.
• B: DynamoDB/DAX would require changing the database model and does not solve the Aurora ingestion design.
• C: SNS is useful for notifications and fanout, but SQS is better for durable buffering and backpressure in this producer-consumer workflow.

• Correct answer: D) Microservices using API

• Why:

• The Act requires:

• It does not, as a public obligation, mandate publishing a detailed report on the training data used.
• Also, the Act does not specify a fixed “six months” retention for system logs in a way that would clearly make A required; logs for traceability are implied, but the six-month period is not a defined, universal requirement.

• B — Prompts and suggestions retained for 28 days
• C — User engagement data kept for two years

• The key distinction is between content data—your prompts, code context, and generated suggestions—and engagement data, such as usage activity.
• Do not confuse retention with training: enterprise Copilot data is generally not used to train GitHub’s models without appropriate authorization.
• For a current exam, verify the policy version cited by the exam provider. The key’s C is likely correct, but B should be treated cautiously and may no longer reflect the current policy.

• DNS????????????????

• ???????????

• ????????????????

• Context: You’re configuring a Windows Autopilot deployment profile (Profile1) for Windows 11 devices. The hotspot asks you to pick the correct behavior for two statements based on the profile’s settings.

• Box 1: “can modify the desktop settings only for themselves”

• Box 2: “Cortana settings” with options like computer name, keyboard layout, etc.

• Correct choices: A and D.

• Why not B, C, or E?

• Why: The data is unlabeled, so you can’t train a model with known input-output pairs. Unsupervised learning discovers structure in the data by grouping similar customers into clusters or tiers without labels.
• What it does: Clustering and segmentation (e.g., k-means, hierarchical clustering, DBSCAN, Gaussian mixtures) to form meaningful groups.
• Why the other options aren’t as suitable:

• Why: A multi-modal embedding model maps both text and images into a shared vector space, enabling cross-modal similarity and retrieval (e.g., find images matching a text query or text matching an image).

• What it does: Enables efficient similarity search using a vector database or nearest-neighbor indexing.

• Why the other options aren’t suitable:

• Quick implementation note: you’d typically pair the embeddings with a vector store (e.g., ANN index) and possibly fine-tune on domain data for better alignment across modalities.

• A) configure matching hold and delay timers

• B) configure matching key-strings

• C) configure matching priority values

• D) configure unique virtual IP addresses

• Ethical red-teaming is adversarial testing aimed at uncovering weaknesses and potential exploits in an AI system before deployment.
• A (improving accuracy) is not the primary?? of red-teaming.
• B (simulate model risk scenarios) is related but narrower; red-teaming centers on finding vulnerabilities and risk exposure, not just simulating risks.
• D (ensure compliance) is not the main goal of red-teaming.

Simple explanation

• The Azure VPN gateway to connect to
• Authentication settings
• Routes for networks that the client can access

• Batch processing, such as scheduled data transformations
• Streaming processing, such as continuously ingesting events

• Manipulating data with multiple languages: Mainly a benefit of Databricks and Spark.
• Real-time notebook collaboration: A workspace/notebook feature.
• Alerts for query failures: A Databricks SQL or job-monitoring feature.
• Distributing complex operations: Primarily a Spark compute capability.

• A private IP address on the virtual network
• An associated public IP address

• Le rôle n’a pas de clés propres. Des identifiants temporaires (via STS) sont émis lorsque quelqu’un ou quelque chose « assume » le rôle.
• La trust policy déclare qui peut assumer le rôle (par ex. une instance EC2, un utilisateur IAM, ou un service comme Lambda).
• Une fois attaché à une ressource (par exemple comme un instance profile sur une EC2), l’application peut appeler les services AWS avec les permissions du rôle.

• Respect du principe du moindre privilège: on accorde uniquement ce qui est nécessaire.
• Pas de stockages de clés sur les instances; les credentials se renouvellent automatiquement.

• Rôles pour EC2, rôles pour Lambda, accès cross-compte, etc.

• What the scenario tests: How monitoring paths on a Universal Forwarder (UF) behave when you deploy an app via the Deployment Server.
• How monitoring is configured: Log paths to monitor are defined in inputs.conf (within apps/local on the UF). Splunk on the UF merges inputs from all apps and local config.
• Key point: Deploying a new app that adds a monitor for a log file does not typically remove existing monitors. The UF will continue monitoring any previously configured paths and will add the new ones from the deployed app.

• Correct answer: C) /var/log/maillog and /var/log/messages.

• In Azure, the typical cost approach for pay-as-you-go VMs is an Operational Expenditure (OPEX) model. You’re billed for actual usage each month.
• The term “elastic expenditure model” isn’t a standard Azure cost category. Elasticity describes scaling capacity, not the billing category itself.
• So recommending the elastic expenditure model is not correct for this goal.

• The solution does not meet the goal. Answer: No (B).

• Pay-as-you-go aligns with OPEX because costs vary with usage.
• Elasticity is about scaling resources, which affects performance and capacity, not the fundamental cost model.

• IPsec: Protects traffic at the network layer, commonly for VPNs; it is not WPA2’s authentication protocol.
• SSL: Protects specific application sessions and has largely been replaced by TLS.
• SSH: Secures remote command-line administration.

• C — Schema.describeSObjects(new String[]{myObject})[0].fields.getMap()

• D — Schema.getGlobalDescribe().get(myObject).getDescribe().fields.getMap()

• A: sObjectType is not used this way, and fieldSet() is not the appropriate method for retrieving all fields.
• B: mySObject.myObject is not valid dynamic Apex syntax.

• D. Provide public endpoints to store and retrieve data

• E. Manage the infrastructure layer and the operating system

• Classifying data
• Configuring access permissions, such as IAM policies
• Selecting or configuring available encryption options, depending on the requirement

• Answer:

• Why:

• Why the other centers aren’t the right fit here:

• Answer: No (B)

• SaaS (Software as a Service) delivers applications over the internet; you don’t manage the underlying infrastructure or OS.
• Deploying Azure virtual machines requires provisioning and managing virtual machines, OS, updates, and networking—this is classic IaaS (Infrastructure as a Service).
• PaaS is for managed runtime environments (apps hosted on a platform), not for VM deployment.

• Startup scripts run at computer startup, before any user logs on, making them ideal for machine-wide configuration on each new VM.
• They execute under the Local System account, which is appropriate for initial setup tasks.
• Linking a GPO to the OU ensures every new VM in that domain automatically receives the configuration—great for scalability.

• A) SetupComplete.cmd runs only once, right after Windows Setup, not on every boot, so it isn’t ideal for provisioning multiple VMs consistently.
• B) Logon scripts run per user, not per machine, so they aren’t reliable for initial VM configuration before a user signs in.
• D) Placing scripts in a VHD doesn’t automatically execute them.

• Pay-as-you-go in Azure is an operating expense (OPEX) model, not an “elastic” expenditure model.
• Elastic or scalable terms describe capacity planning, not the cost category. The cost should be treated as OPEX since you pay for usage each month.
• The parallel question (Question 12) shows the correct approach is using an operational expenditure model.

• Amazon GuardDuty: Detects suspicious activity and threats using security-related data sources; it is not primarily a software vulnerability scanner.
• Amazon Detective: Helps investigate and analyze the root cause of security findings after an event.
• Amazon Cognito: Provides user authentication and authorization for applications.

• Reusing features across different models and teams
• Consistent feature definitions
• Managing feature versions and metadata
• Using features for both training and inference

• SageMaker Data Wrangler: Prepares and transforms data; it is not primarily a shared feature repository.
• SageMaker Clarify: Helps detect bias and explain model predictions.
• SageMaker Model Cards: Documents models, including intended use, risks, and evaluation results.

• Create and manage catalogs of approved IT services
• Govern and publish infrastructure-as-code (IaC) templates, commonly AWS CloudFormation templates
• Control which users can deploy specific products and configurations
• Apply versioning, permissions, and deployment constraints

• AWS Resource Explorer: Searches and discovers AWS resources; it does not govern IaC templates.
• AWS Organizations: Manages multiple AWS accounts, policies, and consolidated billing.
• AWS Systems Manager: Helps operate and manage deployed resources, such as patching and automation, but is not primarily a catalog for governed IaC products.

• Strongest encryption algorithms: Choose the strongest available encryption and integrity options, typically AES with the largest key size and a secure integrity algorithm.
• Encrypt all traffic: Select tunnel mode rather than transport mode. Tunnel mode encrypts the entire original IP packet, which is appropriate for a site-to-site VPN between branch networks.
• Secret-password authentication: Configure a pre-shared key (PSK) on both VPN concentrators. The key must match at both ends.

• Customers access the portal without needing a full Dynamics 365 license.
• They submit case information through portal forms.
• The submitted data is stored in Dataverse tables, such as the Case table.
• Security is controlled using web roles and table permissions, so customers can access only appropriate records.

• Dynamics 365 Connector — connects apps to Dynamics 365 data but is not the portal’s underlying data platform.
• Microsoft SharePoint — useful for documents and collaboration, not the standard case-management data source for portals.
• Microsoft Azure Storage — stores files or other data, but is not the native portal business-data platform.

• Step 1: From Power BI Desktop, select Get Data, choose Folder, enter the folder path, and click OK, then choose Transform data to open Power Query Editor.

• Step 2: In Power Query Editor, expand the Attributes column.

• Step 3: From the Power Query Editor, select the Content column and use Home > Combine Files (aka “Combine binaries”) to consolidate the files.

• What the dataflow options mean:

• Why C fits best:

• Goal: serving a static website globally with minimal latency.
• Why C works: CloudFront is a Content Delivery Network (CDN) that caches content at many edge locations around the world. When users request the site, CloudFront serves cached copies from the nearest edge location, greatly reducing latency. It also reduces load on your EC2 origin by handling many requests at the edge.
• Why the other options aren’t as good:

• Copilot??????????
• ???????????????
• ??????????
• ?????????????
• ???????????????????

• Prompts and Suggestions: Retained for 28 days
• User Engagement Data: Kept for Two Years