A Generative AI Engineer is deploying a customer-facing, fine-tuned LLM on their public website. Given the large investment the company put into fine tuning this model, and the proprietary nature of…

Question

Options

• AImplement AI guardrails to allow users to configure and enforce compliance
• BLeverage Databricks access control lists (ACLs) to configure permissions for accessing models
• CUse secure model features with Databricks Feature Store
• DApply attribute-based access controls (ABAC) to limit unauthorized access

• A
  9% (3)
• B
  70% (23)
• C
  3% (1)
• D
  18% (6)

Explanation

Model inversion attacks aim to reconstruct proprietary training data or extract model internals by repeatedly querying the model. The most direct mitigation is controlling who can access the model in the first place. Databricks Access Control Lists (ACLs) let you restrict which users, groups, or service principals can call a model serving endpoint-reducing the attack surface by preventing unauthorized parties from ever issuing the probing queries that make inversion attacks possible. Option A (guardrails) addresses output compliance, not access restriction. Option C (Feature Store security) protects feature data, not the deployed model endpoint. Option D (ABAC) is a valid access-control paradigm, but in the Databricks context the concrete mechanism for model endpoint permissions is ACLs, making B the most directly applicable DASF strategy.

No community discussion yet for this question.