NVIDIA NCP-AAI Exam Prep
Agentic AI (Page 5 )

• Ease of exploitation
• Required privileges
• User interaction
• Impact on confidentiality, integrity, and availability

• CVE: Identifies and names a specific publicly known vulnerability; it does not measure severity.
• OSINT: Refers to intelligence gathered from publicly available sources.
• SOAR: Automates and coordinates security operations, such as alert response and playbooks.

• A user connects with their Google account.
• Google associates the user’s SSH key with that identity.
• IAM roles determine whether they can log in.
• Access can be tied to a specific employee for auditing.
• Removing the IAM role removes their login access.

• compute.osLogin: regular OS login access
• compute.osAdminLogin: OS login with administrative privileges, typically including sudo

text 
enable-oslogin=TRUE

• User1 (user1@contoso.com) — This is an external guest in the Fabrikam tenant. Assuming the guest uses email one-time passcode authentication, Azure AD emails a passcode to user1@contoso.com.
• User2 (user2@outlook.com) — An Outlook.com address can authenticate with a Microsoft account, so Azure AD uses that account’s authentication rather than sending an email OTP. The fact that User2 has never accessed Fabrikam resources does not by itself trigger OTP.
• User3 (user3@fabrikam.com) — This is an internal Fabrikam user, so the user signs in with the organization’s normal credentials and does not receive a guest OTP.

• 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.

• Filtering harmful or inappropriate content
• Blocking specified topics
• Detecting and masking sensitive information
• Enforcing content policies suitable for children

• Amazon Rekognition analyzes images and videos; it does not control text generated by a language model.
• Amazon Bedrock playgrounds are for testing models and prompts, not enforcing production safety policies.
• Agents for Amazon Bedrock help models complete tasks and use tools; they do not primarily provide content-safety controls.

• 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.

• Assign endpoints to a group based on the result of one policy.
• Use that group as the scope of another policy.
• This effectively passes the endpoints from one policy’s processing flow to another.

• 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.

• 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.

1. Product + ProductSubCategory: Inner join

1. Result + ProductCategory: Left outer join

• Continuous deployment: Automates software releases; it does not specifically test input handling.
• Static code analysis: Examines source or compiled code without executing it. It can find coding flaws but does not directly stress the application with malformed inputs.
• Manual peer review: Developers inspect code or design manually. Useful, but not the best practice specifically for testing how a live product handles invalid input.

• Maximum response and resolution times
• Expected uptime or availability
• Performance requirements
• Escalation procedures
• Consequences for failing to meet requirements

• SOW: Defines a project’s scope, tasks, deliverables, and schedule.
• MOA: Documents specific responsibilities agreed upon by two parties.
• MOU: Records a general understanding or intent to cooperate.

• The ERSPAN source is the device or interface whose traffic is being copied.
• The ERSPAN destination is the device receiving the encapsulated copy and forwarding it to the traffic analyzer.
• The local monitor session number is locally significant. Therefore, the source and destination do not need matching session IDs such as 47 or 48.
• The erspan-id identifies the ERSPAN stream and must correspond between the source and destination configuration.

• vBond — orchestration plane: Authenticates devices, helps them discover each other, and assists with NAT traversal and initial control-connection setup.
• vSmart — control plane: Distributes routing and policy information to WAN Edge devices using the overlay control plane.
• vManage — management plane: Provides centralized configuration, monitoring, and orchestration through its GUI/API.
• WAN Edge — data plane: Forwards user traffic across the SD-WAN fabric.

1. Write a structured log entry to CloudWatch Logs containing:

1. Create a CloudWatch Logs metric filter that:

text 
ApiRequests 
Operation=GetOrders 
ResponseCode=500 
Version=2.4.0

• B is incorrect: CloudWatch Logs Insights queries analyze logs but do not directly publish metrics in this manner.
• C is incorrect: ALB access logs do not contain arbitrary Lambda response metadata as custom fields.
• D is incorrect: X-Ray traces and insights are not the appropriate mechanism for creating these custom, dimensioned application metrics.

• A password cracker would normally be identified by behavior such as loading password hashes or attempting dictionary/brute-force attacks.
• The question describes a routine audit, not password recovery.
• Simply encountering a protected file during an audit is characteristic of a misconfigured or insufficiently privileged FIM.

• NIDS: Monitors network traffic for suspicious activity and network-based attacks targeting the application or its supporting services.
• HIPS: Protects an individual host by monitoring processes, files, and system activity. It can help secure the web server, but it is not focused specifically on the application’s network attack surface.
• SIEM: Aggregates and correlates logs; it does not directly prevent attacks.
• Honeypot: A decoy system, not the primary protection for a production web application.
• Certificate revocation list: Deals with certificate trust, not application exploitation.

• Microservices: An application design approach that divides software into small services; it does not specifically reduce the number of operating systems.
• Containerization: Uses isolated containers that share the host OS kernel—best match.
• Virtualization: Creates separate virtual machines, and each typically requires its own guest OS, so it does not minimize OS instances as effectively.
• Infrastructure as code: Automates deployment and management of infrastructure; it does not describe the architecture being used.

• C — Some risks cannot be eliminated.

• D — Accepting residual risk does not imply recklessness.

• E — Most companies in this industry will say something similar.

• Authentication — verifies who the user is.
• Authorization — determines what the authenticated user is allowed to access.
• Accounting — records session activity, such as login time and resource usage.

• CIA — confidentiality, integrity, and availability; core security objectives.
• ACL — access control list; rules that permit or deny traffic or access.
• PEM — Privacy Enhanced Mail, an older email security standard.

• AWS AppSync is designed for serverless applications that need to retrieve and combine data from multiple sources.
• A pipeline resolver can invoke multiple functions or data-source resolvers, including resolvers connected to different DynamoDB tables.
• The application can receive the combined result through one GraphQL request, without requiring custom aggregation logic in the application.
• This is operationally efficient because AppSync manages the API layer and resolver execution.

• Athena Federated Query is intended primarily for analytics and ad hoc querying, not low-latency application data retrieval.
• Querying DynamoDB through Athena can consume DynamoDB read capacity and may affect table performance.
• It also introduces analytical-query overhead and is not the normal choice for a microservice web application.

• 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. Get Data > Folder

1. Expand the Attributes column

1. Remove the Content column

• Folder connector: lists files and their metadata, but also includes a Content binary column.
• Combine Files: processes and combines the actual CSV contents, which is explicitly disallowed here.

1. The specific table — for example, incident
2. The parent table — for example, task
3. The wildcard table — *, which applies broadly

• The application failing to start or becoming unhealthy
• An incorrect probe path, port, or protocol
• The application needing more startup time
• Dependency or configuration failures
• A liveness probe being too aggressive

bash 
kubectl describe pod <pod-name> 
kubectl logs <pod-name> --previous

• Scale to more replicas: may hide the issue but does not diagnose it.
• Decrease initialDelaySeconds: makes probes run earlier and could worsen startup failures.
• Drain and reboot the node: node metrics are normal, so rebooting is not justified as the first step.

• Accept: Acknowledge the risk and take no additional action.
• Transfer: Shift the risk or financial impact to another party, such as through insurance or a contract.
• Mitigate: Reduce the likelihood or impact using controls, such as firewalls, training, or backups.
• Avoid: Eliminate the activity that creates the risk.

1. You try to access a cloud application, such as accounting software.
2. The application redirects you to the company’s IdP.
3. You authenticate with the IdP.
4. The IdP sends a digitally signed SAML assertion to the application.
5. The application validates the assertion and grants access.

• The user’s identity
• Whether authentication succeeded
• Roles or group membership
• Access-related attributes

• Business
• People
• Governance
• Platform
• Security
• Operations

• Cloud financial management
• Portfolio management
• Data governance
• Risk management
• Compliance

• STP elects the switch with the lowest Bridge ID as the root bridge.
• The Bridge ID consists mainly of bridge priority + MAC address.
• If multiple switches have the same priority, the switch with the lower MAC address is preferred.
• STP bridge priority values are configured in increments of 4096, such as 0, 4096, 8192, and so on. 32768 is the standard default value.

• Honeypots: Decoy systems or services made to attract attackers. Legitimate users should not interact with them, so activity there is a strong indicator of malicious behavior.
• DNS sinkholes: Redirect requests for known malicious domains to a controlled internal system. This can reveal infected hosts attempting to contact command-and-control infrastructure.

• Tokenization: Replaces sensitive data with non-sensitive tokens.
• CI/CD: Automates software development, testing, and deployment.
• Threat modeling: Identifies and analyzes potential threats before or during system design.
• Data obfuscation: Makes data harder to understand or interpret.

• 10 TB with maximum I/O performance: Use EC2 instance store. It provides very high-throughput, low-latency local storage suitable for temporary video-processing data. The application must tolerate data loss because instance store is ephemeral.
• 300 TB of highly durable media: Use Amazon S3, which is designed for highly durable object storage and can scale to this size.
• 900 TB of inactive archival media: Use S3 Glacier, which provides lower-cost storage for data that is rarely accessed.

• A: Amazon EBS offers high performance but is not the best fit for the question’s “maximum possible I/O” requirement compared with local instance store.
• B: Amazon EFS is durable and shared, but is generally more expensive and less suitable for very large archival storage.
• C: Uses EC2 instance store for performance, but incorrectly uses EFS for the 300-TB durable tier and S3 rather than Glacier for archival data.

• Amazon S3 as the origin for static content
• Application Load Balancer (ALB) as the origin for dynamic content
• Amazon Route 53 configured to point the company’s domain name to CloudFront

• /static/* ? S3 origin; cache content at edge locations
• Dynamic paths such as /api/* ? ALB origin; forward requests to the EC2 application

• Server provisioning
• Operating systems and runtime infrastructure
• Scaling
• Much of the maintenance and patching

• Thin clients reduce endpoint hardware requirements but do not specifically streamline code deployment.
• Private cloud provides dedicated cloud infrastructure but usually requires more management and expense.
• Virtual machines improve resource utilization, but administrators still manage the VMs, operating systems, and deployments.

• Suspended an operator who correctly refused to use unsafe equipment.
• Rewarded another operator for using the equipment despite the broken safety guard.

• A — Management is not particularly concerned with managing risks.

• B — Employees will perceive that management does not care about health and safety risks.

• C — This factory’s safety record is likely to be poor.

1. x = -100
2. -100 < x < 100
3. 100 = x < 1000
4. x = 1000

• -500 ? class 1
• 0 ? class 2
• 100 ? class 3
• 1000 ? class 4

• SQL injection
• Cross-site scripting (XSS)
• Malicious file uploads
• Common web-application exploits

• Layer 4 firewall: Filters primarily by IP addresses, ports, and transport protocols; it lacks detailed web-application inspection.
• NGFW: Provides advanced network filtering and may include application awareness, but a WAF is more specialized for protecting web applications.
• UTM: Combines several security functions in one device, but it is not specifically focused on web-application attacks.

• The type of change: standard, normal, or emergency
• The risk, impact, and complexity
• The change model being used

• A standard change may be pre-authorized through its change model.
• A normal change may require assessment and approval by an appropriate change authority.
• An emergency change may be authorized by an emergency change authority, often using an expedited process.

• Change authorities are not only needed for emergency changes.
• They are not assigned only when a change is deployed.
• They are not only required for normal changes.

• Storage usage of files stored in Microsoft Teams: SharePoint site usage report

• Number of active users per team: Microsoft Teams usage report

1. SharePoint site usage report
2. Microsoft Teams usage report

• Internal interface ? connected toward the corporate network/internal firewall
• External interface ? connected toward the Internet-facing firewall

1. A route for the internal corporate network through the internal-facing interface/firewall.
2. A default route (0.0.0.0/0) through the Internet-facing interface/firewall.

• Online payment fraud
• Account takeover
• Fake account creation
• Suspicious transactions

CIMA P3 topics

• IT governance and systems strategy
• Systems development and implementation
• Change and project governance
• Strategic and business risk

• A — Plan for new systems: The committee should help identify and prioritise systems requirements and ensure planned systems support organisational needs.
• D — Consider competitive issues: New information systems may affect Y’s competitive position, service quality and efficiency. These strategic implications should be considered.
• E — Ensure the system meets company goals: This is a key governance responsibility. The committee should ensure the project remains aligned with Y’s objectives and delivers business value.

• B — Develop new systems: Detailed system development should be performed by IT specialists and project teams, not the steering committee.
• C — Manage the project: The project manager should handle day-to-day project management. The steering committee should provide oversight, challenge and key decisions, rather than manage daily activities.

• desktopwidth:i:1280
• desktopheight:i:1024

• Connection information: Controls connection-related settings, not the desktop resolution.
• Scaling plan: Controls when session hosts start, stop, or scale based on demand.
• Session behaviour: Controls behavior such as disconnect or logoff handling.
• Display settings: Specifically controls RDP display properties, making it the appropriate choice.

• Third-party enterprise application with embedded AI: The provider manages most security controls. The customer mainly manages users, data, and configuration.
• Application using an existing third-party FM: The customer controls the application and its inputs/outputs, but the provider controls the underlying model and infrastructure.
• Fine-tuning a third-party FM: The customer has additional responsibility for training data, data protection, and model behavior, but still depends on the provider’s base model and platform.
• Building and training a model from scratch: The customer controls the model, training data, application, deployment, and many security safeguards. This creates the greatest security ownership.

• Create and manage approved IT services as products
• Control which users can deploy those products
• Apply governance, permissions, and constraints
• Use and manage infrastructure-as-code templates, such as AWS CloudFormation templates

• AWS Resource Explorer: Helps find and browse AWS resources; it does not govern IaC deployments.
• AWS Organizations: Manages accounts, consolidated billing, and organization-wide policies.
• AWS Systems Manager: Provides operational management, such as patching, automation, and parameter storage, but it is not primarily used to catalog and govern IaC-based service deployments.

• Documented and evaluated
• Assessed for security, cost, and operational impact
• Reviewed and approved by the appropriate personnel
• Tested before implementation
• Tracked with version history and rollback procedures

• Risk management process: May help assess the issues, but it does not authorize and manage design changes.
• Product design process: Applies primarily while creating a new design.
• Design review process: Normally evaluates a design before release; this design is already released.

• Reviewing security objectives to ensure they support business goals.
• Ensuring security requirements are integrated consistently across business units.
• Coordinating security policies, risk management, controls, and awareness activities.

• Executive management: Provides direction, funding, and support, but usually does not manage day-to-day security integration.
• Board of directors: Provides governance and oversight, not operational security coordination.
• Steering committee: Helps coordinate and prioritize initiatives, but the CISO typically owns the overall security program and ensures its integration.

• Have a definite, clear objective (C): Know exactly what you want to communicate. A clear purpose reduces ambiguity and irrelevant information.
• Plan the communication (E): Decide what to say, who needs the information, the appropriate channel and the best timing.
• Seek feedback (F): Check that the receiver understood the message. This could involve asking for confirmation or inviting questions.

• Ask questions (A): Usually associated with the receiver checking or clarifying information, although a sender may use questions in practice.
• Give feedback (B): Feedback is generally the receiver’s response to the sender’s message. The sender’s responsibility here is to seek feedback.
• Consider their contribution (D): This is not a standard sender-focused step in the communication process. The sender should consider the receiver’s needs and perspective, but the wording is less precise than the three recognised actions above.

• A — Capacity: The system must handle all current records and anticipated future records. A system that cannot cope with volume will quickly become ineffective.
• C — Accuracy: Academic records are critical, so the new system must be at least as accurate as the existing system and comparable systems. Poor accuracy could cause serious administrative and reputational problems.
• D — Integration: Linking with the university’s financial systems supports consistency and avoids duplicate data entry, especially for registration, deregistration and fee-related transactions.

• B — Parallel running: This is a useful implementation and risk-control technique, but it is not an essential feature of the system itself. It may also be expensive and time-consuming.
• E — Internet access for all information: This creates significant confidentiality and security risks. External access might be needed for selected services, but unrestricted access to all information is not essential and would be inappropriate for student records.

• transform() — modifies every element in an array
• filter() — keeps elements matching a condition
• exists() — checks whether any element matches
• aggregate() — reduces array elements to one result

sql 
SELECT transform(scores, x -> x * 2) 
FROM student_scores;

• A is incorrect because higher-order functions are mainly useful for nested or complex data.
• B is incorrect because they do not convert logic into Python.
• D is incorrect because their purpose is not to fix slow built-in functions, but to process arrays or maps with custom logic.

• Cloud architecture
• Data architecture
• Platform engineering
• Data engineering
• Provisioning and modernization

• Data protection is associated mainly with security and risk management.
• Data governance belongs to the governance perspective.
• Data science is not an AWS CAF capability.