Secure Model Storage and Access Controls

1. In zero-trust security architecture for AI systems, what principle is emphasized?

Trust all internal network traffic

Never trust, always verify every access request

Trust users but not machines

Verify only external access

2. In compliance with NIST Cybersecurity Framework, what does the 'Protect' function require for AI models?

Only detecting breaches after they occur

Implementing access controls, encryption, and data security measures

Eliminating all security measures

Trusting all users without verification

3. What is the primary security advantage of using a dedicated model serving gateway?

It eliminates the need for authentication

It centralizes access control and monitoring

It increases model training speed

It reduces storage requirements

4. Which cryptographic hash function is most suitable for verifying model integrity?

MD5 or SHA-1

SHA-256 or SHA-512

Base64 encoding

Simple checksums

5. In secure model access, what does principle of least privilege (PoLP) mean?

Grant all users full access to all models

Users get only the minimum permissions needed for their role

Eliminate all access controls

Share passwords across teams

6. What is the purpose of implementing network segmentation for model storage infrastructure?

To increase model accuracy

To isolate sensitive models from untrusted networks

To reduce storage costs

To speed up data transmission

7. Which threat specifically targets the integrity of stored AI models through unauthorized modification?

Denial of service

Model poisoning

SQL injection

Cross-site scripting

8. In secure model access, what is the primary function of a model registry?

To train new models automatically

To catalog, version, and control access to models

To increase model inference speed

To eliminate the need for encryption

9. What does a digital signature verify about a stored AI model?

The model's inference speed

The model's origin and that it hasn't been tampered with

The model's accuracy score

The training dataset size

10. Which storage solution provides the highest availability and redundancy for critical AI models?

Single local hard drive

Geo-replicated cloud storage with backups

USB external drive

Unencrypted network share

11. Which encryption standard is most commonly used to protect AI models at rest in enterprise systems?

AES-256

DES

RC4

MD5

12. What is the primary benefit of using model containerization with signed images?

It eliminates all security risks

It ensures model integrity and authenticity

It speeds up model training

It reduces model size significantly

13. Which of the following is a best practice for managing API keys used to access stored models?

Store keys in plain text in configuration files

Rotate keys regularly and use secret management systems

Share keys across all team members

Embed keys directly in source code

14. What does TLS/SSL encryption primarily protect during model transmission?

Storage disk space

Data in transit between client and server

CPU processing speed

Model training time

15. In secure model storage, what is the purpose of audit logging?

To improve model accuracy

To track all access and modifications to models

To encrypt data automatically

To reduce storage requirements

16. Which compliance framework specifically addresses data protection and privacy in EU member states?

HIPAA

GDPR

SOC 2

NIST

17. What does a hardware security module (HSM) primarily protect in model storage systems?

Network bandwidth

Cryptographic keys

Model inference latency

User interface design

18. Which authentication method provides the strongest protection for API access to stored models?

Username and password only

Multi-factor authentication (MFA)

IP whitelisting alone

API keys without expiration

19. In model versioning systems, what does immutability ensure?

Models load faster

Previous versions cannot be altered or deleted unexpectedly

Models are automatically updated

Storage costs decrease

20. What is the primary purpose of role-based access control (RBAC) in AI model repositories?

To encrypt all data automatically

To restrict model access based on user responsibilities