ISACA AAIA Exam Prep
ISACA Advanced in AI Audit (Page 49 )

• Approve major information security policies
• Allocate funding and resources
• Set the organization’s risk appetite
• Ensure security supports business objectives

• Evaluating vendors is generally handled by security, procurement, or technical teams.
• Assessing risks is typically performed by security and risk-management professionals, with management accepting or prioritizing the risks.
• Monitoring regulatory adherence is usually handled by compliance, audit, or security functions.

• CSIRT: A team that responds to and manages security incidents.
• CVSS: A scoring system for rating the severity of vulnerabilities.
• SOAR: A technology platform for automating and coordinating security operations and incident response.

• VM CPU, memory, disk, and network usage
• Storage transactions and latency
• App Service response time
• Database DTU or CPU utilization

• Azure Traffic Analytics analyzes network traffic flows and security patterns; it is not the general infrastructure metrics tool.
• Azure Activity Log records subscription-level control-plane operations, such as resource creation, deletion, or configuration changes.
• Azure Advisor provides recommendations about cost, security, reliability, performance, and operational excellence, but it is not the primary tool for investigating live metrics.

• A: Reduce management overhead — This is generally an advantage of public cloud, not hybrid cloud. Hybrid environments can increase management complexity.
• B: Cross-platform virtualization — Not the defining advantage of hybrid cloud.
• D: Leverage IaaS and PaaS — Public clouds commonly provide both IaaS and PaaS, so this is not unique to hybrid cloud.

• The servers run different operating systems: Windows, Linux, and Unix.
• Containers generally share the host operating system’s kernel, so they cannot normally run workloads requiring unrelated kernels, such as Windows and Linux, on the same Linux host.
• A hypervisor can create separate virtual machines, allowing each VM to run its own guest OS.
• A Type 2 hypervisor runs on top of a host operating system and can support multiple guest operating systems.

• Disable default accounts: Default or unused accounts may have predictable names or credentials and can be abused by attackers. They should be disabled, removed, or secured with strong credentials where necessary.
• Remove unnecessary services: Every running service can contain vulnerabilities or provide an entry point. Removing services the server does not need reduces exposure.

• Add the server to the asset inventory: Important for asset management, but it does not directly harden the system.
• Document default passwords: Default passwords should be changed or disabled—not recorded for continued use.
• Send logs to the SIEM: Supports monitoring and detection, rather than directly reducing the attack surface.
• Join the corporate domain: A deployment/configuration decision, not inherently a hardening measure.

• Whether a user has a console password
• Password age and last password change
• Whether MFA is enabled
• Access key status
• Access key creation date and last-use information
• Access key age, which helps audit rotation

• IAM Access Analyzer identifies unintended access to resources; it does not report credential age or rotation.
• AWS Artifact provides AWS compliance documents and agreements, not details about your IAM users.
• AWS Audit Manager helps collect evidence for audits, but it is not the direct tool for viewing IAM credential rotation details.

• R-squared: Typically used for regression, not classification.
• RMSE: Measures prediction error for continuous numerical values, so it is not the best fit here.
• Learning rate: A training hyperparameter that controls how quickly the model updates its weights; it is not an evaluation metric.

• 16 VMs: one failure loses 25 GB and 10 cores—only 6.25% of memory.
• 8 VMs: one failure loses 12.5% of memory.
• 4 VMs: one failure loses 25%.
• 1 VM: one failure loses the entire application environment.

• You can configure different pre-shared keys for different device groups.
• All cameras could use one key.
• All scanners could use another key.
• Each key can be associated with an appropriate role or access policy.

• An active relationship is used automatically when filters are evaluated. An inactive relationship does not propagate filters unless explicitly activated in DAX, so it would not reliably enforce the required security behavior.
• Apply security filter in both directions allows RLS filters to propagate across the relationship in both directions. This ensures that bidirectional filtering respects the RLS rules applied to both tables.
• Assume referential integrity is unrelated to RLS. It is mainly a DirectQuery optimization that allows Power BI to use an inner join when matching keys are guaranteed to exist.
• The relationship must therefore be active, with security filtering enabled in both directions.

• Restarting the device: May fix software glitches, but would not usually make cable orientation matter.
• Swollen battery: A safety concern, but it would not typically cause recognition only in one cable orientation.
• Failing cable: Possible, but less likely because the problem occurs with this particular device and not other devices of the same type.

• Encryption to preserve confidentiality
• Integrity protection to detect alteration
• Authentication of the server through certificates

• SHA-256: A hashing algorithm; it does not encrypt data for transmission.
• SSL 3.0: Obsolete and insecure. It should not be used.
• AES-256: A strong encryption algorithm, but by itself it does not provide the complete secure communication protocol needed for data in transit.

• Restrict a port to a defined number of MAC addresses
• Use statically assigned or “sticky” MAC addresses
• Shut down or restrict the port when violations occur

• Load balancer: Distributes application traffic; it does not protect switch MAC tables.
• IPS: May detect suspicious network activity, but is not the primary control for limiting switch MAC entries.
• NGFW: Filters traffic at the network boundary; it does not directly prevent MAC table flooding on access switches.

• C series: CPU-optimized instances; not designed primarily for LLM training.
• G series: GPU instances generally optimized for graphics and some machine learning workloads, but not specifically large-scale model training efficiency.
• P series: Powerful GPU instances for ML training, but typically consume more energy than Trainium-based instances for suitable training workloads.

1. Convert text and images into numerical embeddings.
2. Store or compare those embeddings.
3. Retrieve the most semantically similar items.

• Embedding model: Best for similarity search and retrieval.
• Generation model: Best for producing new text or other content from multimodal input.

1. Ensure BOMs can include subcomponents ? Configure an entity relationship

1. Report changes made by engineers ? Enable auditing

• Configure entity relationship
• Configure auditing

Statement 1 — Correct

Statement 2 — Correct

• Share of overall contribution: part of the organization’s total contribution is allocated to the supplying division.
• Dual pricing: the supplying division may record a higher transfer price while the receiving division records a lower price, with central accounts absorbing the difference.
• A centrally imposed price range: this may allow the transfer price to be set above variable cost, helping the supplying division recover fixed costs.

• principal represents the source or initiating system.
• principal.port is the source port.
• target represents the destination system.
• target.port is the destination port.

text 
principal.port = 5555

• network.ApplicationProtocol = HTTP identifies HTTP traffic but not the backdoor specifically.
• target.port = 5555 finds traffic going to port 5555, rather than originating from it.
• network.ip_protocol = TCP is too broad because many services use TCP.

• An Azure pay-as-you-go subscription follows an operational expenditure (OpEx) model.
• With OpEx, the organization pays for services as they are used, typically based on consumption.
• There is no need to purchase and maintain physical servers upfront.
• This contrasts with capital expenditure (CapEx), where the company makes a significant upfront investment in infrastructure, such as on-premises Hyper-V hosts.

• Data Category Groups e categorias;
• Visibilidade dessas categorias para perfis ou permission sets.

• B e C: regras de compartilhamento não são o mecanismo apropriado para restringir resultados de busca do Knowledge por divisão.
• D: a role hierarchy não é o mecanismo principal para controlar a visibilidade de categorias do Knowledge. Além disso, a configuração correta é atribuir visibilidade a perfis ou permission sets.

• Portfolio definition cycle — establishes and maintains the portfolio strategy, objectives, and management framework.
• Portfolio delivery cycle — coordinates and monitors the delivery of change initiatives, ensuring the portfolio remains aligned with strategic objectives and delivers value.

• Practice — a structured area of portfolio management activity, such as risk management or resource management.
• Technique — a method used to perform an activity; it is not the term for portfolio delivery itself.
• Principle — a guiding rule for effective portfolio management.

• Domain controllers, because the requirement specifically excludes them. However, note that both member servers and domain controllers run Windows Server, so OS targeting alone cannot distinguish them unless the targeting option allows selecting the relevant server role/version. In the usual exam interpretation, target the applicable server OS category.
• Client computers, because they run a client Windows operating system.

• Domain: Targets membership in a domain, not the computer’s operating-system type.
• Security Group: Would require creating and maintaining a group containing all member servers, increasing administrative effort.
• Environment Variable: Not designed to reliably identify member servers.

• Sensitive document shared during a Teams chat: Teams chat
• Sensitive document stored/shared in a Teams channel: Teams channel messages

• Files shared in a chat are generally stored in the sender’s OneDrive, but the relevant DLP context is the Teams chat message and its attachment.
• Files shared in a channel are stored in the team’s SharePoint site, but the relevant context is the Teams channel message.
• The DLP rule can then use an action such as restrict access to the content to prevent guest users from opening the sensitive document.

• Code suggestions: Inline completions as you write code, such as lines, functions, or blocks.
• Chat: An interactive interface for asking questions, explaining code, generating code, and helping with debugging.

• Knowledge Base: Not listed as a standard Copilot Individual IDE feature in this context.
• Pull Request Diff Analysis: More closely associated with GitHub-based code review or pull-request features, not the core Individual IDE extension capabilities.

Question 3: AWS Storage Gateway File Gateway

• A: S3 bucket per user — S3 is object storage, not a traditional shared file system; mounting it requires extra tools and is less efficient.
• B: File Gateway — Provides a local file share using standard file protocols while storing files in Amazon S3. This preserves convenient local access and extends storage capacity.
• C: WorkSpaces and WorkDocs — Designed for virtual desktops and document collaboration, not large shared file storage.
• D: EC2 with EBS — EBS is attached to one EC2 instance and is not designed for direct multi-user file sharing.

• Default network device credentials: More associated with external attackers or poor device configuration.
• Spear phishing emails: Commonly used to compromise users, but less characteristic of an insider who already has access.
• Typosquatting and impersonation: Primarily social-engineering techniques aimed at deceiving users.
• Unidentified removable devices: Provide a direct way to transfer data outside the organization and may bypass network monitoring.

• D — Create sibling resource pools for the Production VMs and Test VMs.

• B — Configure a memory reservation on the Production resource pool.

• Shares only determine relative priority when resources are scarce; they do not guarantee memory.
• A parent resource pool alone does not provide a guarantee.
• A child pool for Test VMs does not reserve memory for Production.

• Inherent risk: Risk that exists before controls are implemented.
• Avoided risk: Risk eliminated by choosing not to perform the risky activity.
• Operational risk: Risk arising from failures in processes, people, systems, or daily operations.

• Economy: Obtaining inputs at the lowest appropriate cost.
• Efficiency: Making the best use of resources, such as staff, money, and equipment.
• Effectiveness: Achieving the intended objectives or outcomes.

• Each team’s Bedrock workflow uses a separate IAM service role.
• The role’s Amazon S3 permissions can be limited to only that team’s customer folder or objects.
• This enforces least privilege and prevents one team’s Bedrock application from accessing another team’s data.
• Access is enforced by AWS permissions, not by asking users to provide a customer name in each request.

• A single role with broad S3 access does not provide team-level isolation.
• Asking teams to specify a customer name is not an access-control mechanism.
• Redacting data does not prevent unauthorized access.
• A bucket policy alone is insufficient unless it explicitly enforces the required team-role restrictions.

• Generates a syslog message
• Shuts down the PoE port
• Places the port into error-disabled (err-disabled) state
• Frees the power allocation

• A device starting to draw power does not necessarily generate a syslog message.
• Monitoring does not pause data traffic while checking power.
• Drawing less than the configured minimum does not automatically mean the device has failed and should be disconnected.

• AWS Network Firewall can inspect and block traffic across a VPC when traffic is routed through the firewall. A Drop action rule can deny traffic from the suspicious source IP.
• GuardDuty findings are sent to Security Hub.
• A Lambda function can be invoked when Security Hub receives a new GuardDuty finding, extract the source IP, and update the Network Firewall policy with a blocking rule.

• GuardDuty threat lists: Used to identify or increase the severity of findings; they do not directly block VPC traffic.
• AWS WAF: Protects supported web applications, such as CloudFront or Application Load Balancers. It does not provide general VPC-wide traffic blocking.
• GuardDuty suppression rules: Suppress or hide findings. They do not block the suspicious traffic.

• Documented and formally requested
• Analyzed for security, cost, and operational impact
• Reviewed and approved by the appropriate personnel
• Tested before implementation
• Tracked and reversible if problems occur

• Risk management process: Identifies and evaluates risk, but does not directly authorize design changes.
• Product design process: Applies primarily while creating or developing the product.
• Design review process: May have identified the issues, but the next step for changing an already released design is controlled change management.

• A — Custom Button: A button can launch a flow, typically by passing the flow’s URL and record context.
• D — From a Process: A Process can invoke a flow using an Apex action configured for flow invocation. However, this is automation-triggered, not directly accessed by a user.
• B — Quick Action: In current Salesforce functionality, a Flow Quick Action can launch a screen flow directly from a record page. This makes B a valid modern method.
• C — Visual Workflow Launcher: This is not generally listed as a standard access method in current Salesforce terminology; Salesforce commonly uses the Flow component on Lightning pages instead.

• Microservices: An application-design approach that divides an application into smaller services. It does not inherently reduce the number of operating systems.
• Virtualization: Creates virtual machines, but each VM generally requires its own complete operating system. This improves isolation but does not minimize OS instances as effectively.
• Infrastructure as code: Uses code to automate infrastructure deployment and configuration. It improves consistency but does not determine how many operating systems are used.

1. Define Power Platform DLP policies

1. Add custom prompt modifications

1. Conduct adversarial testing

• Tokenization: Replaces the value with a nonsensitive token, often reversible through a secure token vault.
• Hashing: Converts data into a one-way fixed-length value, generally used for integrity checks or password storage.
• Obfuscation: A broad term for making data difficult to understand; it is less precise than masking in this context.

• Authentication: Verifies a user’s identity, such as username and password.
• Authorization: Determines what the authenticated user is allowed to access.
• Accounting: Records connection and usage details, such as login time or session duration.

• CIA: Confidentiality, integrity, and availability are broad security objectives, not a specific RADIUS function.
• ACL: An access control list contains permit/deny rules on systems or network devices. RADIUS can support access decisions but is not itself an ACL.
• PEM: Privacy-Enhanced Mail is also associated with certificate and key encoding formats; it is unrelated to RADIUS’s primary purpose.

• Planning and designing products and services
• Developing and delivering them
• Managing risks, suppliers, and resources
• Measuring performance
• Continually improving value delivery

• Value stream steps describe activities used to create and deliver value; they do not themselves provide the broader capabilities.
• Vision and operating model provide direction and structure, but they are not the practical management capabilities that enable the work.
• Value chain is the operating model showing how activities combine to create value; it is enabled by practices and resources.

• 800 devices ? a scalable, centralized solution is needed.
• Four hardware/vendor classes ? the tool must handle different drivers and images.
• Two user groups ? it should allow different configurations for each group.
• “Distribute custom images” ? an imaging/deployment tool is required, not just a file-copy utility.

• Separate accounts/domains create a strong administrative boundary between non-production and production.
• Separate projects provide distinct IAM boundaries, billing, quotas, service accounts, and resources.
• Developers can receive permissions in the development project without automatically gaining access to production.
• Production administrators can be restricted from development resources, supporting separation of duties and least privilege.

• A: Creates excessive project fragmentation without being required by the stated requirement.
• C: Keeps development and production under one account and does not clearly separate administrative duties.
• D: Uses one account and only two projects; it provides some project isolation but weaker organizational separation.

• The requirement is to obtain an assessment of the Azure environment’s design.
• The proposed solution is the Premier support plan.
• Premier can provide this type of guidance, but it is a higher-tier, more expensive option.
• The question specifically says to keep costs to a minimum. A lower tier—typically Professional Direct in the support-plan model used by this question—provides architectural/design guidance.

• The S3 bucket as an origin for static content
• The Application Load Balancer (ALB) as an origin for dynamic content
• An Amazon Route 53 alias record pointing the application domain to CloudFront

Why this works

• Static data: CloudFront caches objects from S3 at edge locations, reducing latency for users worldwide.
• Dynamic data: CloudFront forwards requests that require application processing to the ALB. It can also maintain persistent connections and use the AWS global edge network to improve delivery performance.
• One domain: CloudFront behaviors can route different path patterns to different origins—for example, /static/* to S3 and other paths to the ALB.

Why the other options are incorrect

• Global Accelerator is designed primarily for accelerating traffic to applications with endpoints such as ALBs, not for caching S3 content.
• Using separate domains or splitting static and dynamic traffic adds unnecessary complexity.
• CloudFront already provides the required global edge distribution and caching, so combining it with Global Accelerator is unnecessary here.

• Supervised learning requires labeled examples to learn how to classify customers.
• Reinforcement learning learns through actions, rewards, and penalties, which is not the main task here.
• RLHF uses human feedback to improve model behavior, typically for generative AI, not customer segmentation.

• Check whether the data is accurate, complete, and properly prepared.
• Confirm that the pattern is not caused by an outlier, data-quality problem, or modeling issue.
• If the data is valid, treat the result as a potentially valuable new business insight.

• A: A data scientist may help, but expert validation and data-quality checking come first.
• C: Filtering out inconvenient data risks hiding a valid insight.
• D: Einstein’s result should not be accepted without verification simply because it produced the finding.

• 2000::/3 — Global unicast, used for publicly routable IPv6 addresses.
• FC00::/7 — Unique local, used for private internal addressing.
• FE80::/10 — Link-local, used only on the local network segment and not routed between subnets.

• Normalize casing in the data source—for example, convert all values to uppercase or lowercase.
• Add an index key to provide a unique, stable identifier for each row or value.