Applying Zero Trust to your Artificial Intelligence (AI) System Development
In government networks, applying CISA’s Zero Trust principles with the NIST Artificial Risk Management Framework (NIST AI RMF 1.0) crafts a powerful security strategy for your AI systems. In this digital transformation era, where AI technologies play a central role across various sectors, ensuring the security and integrity of our data that feeds AI is paramount.
In this blog, I’m focusing on the AI RMF MAP core function and aligning with CISA’s Data Pillar to provide a quick start in assisting on ensuring a Zero Trust approach to your AI endeavors. Future blogs will cover the Govern anchor for the AI RMF core as well as the measure and manage functions.
Map: “Context is recognized and risks related to context are identified”
Although you need to work through the specifics of your AI system and most importantly your data sources to protect in your AI systems, the CISA Zero Trust Maturity Model 2 gives us a head start on approaching the mapping.
The two major functional concepts from the CISA Data Pillar and summarized for this blog are:
Data - inventory, categorization, availability
Data - protection and controls
For this example, we will go deeper on the protection and controls which will help you map your protection and show the aligning between ZT and AI RMF.
Data - Protection and Controls Snippet from Risk Catalog
Here are some sample risks that would be considered on an AI system that provides a web interface for user interaction and an API for the data query:
Risks:
Unauthorized access via AI APIs or AI BPIs API. Errors or manipulation of API provides direct access to agency data bypassing functional (safe) queries
Inline data points on premise or cloud can intercept protected traffic with network monitorin
Possible Controls Solutions: API or BPI application level fireballing (native), or Standalone Application Level firewall, or LOAD balancers with API policy.
Summary
This alignment, at AI design kickoff will give you lift across both the NIST AI RMF and the CISA Zero Trust Maturity Model 2 and avoiding having to bolt on security later.
I will cover the other Core functions of the NIST AI RMF in future blogs.
References:
NIST Artificial Intelligence Risk Management Framework (AI RMF 1.0): https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf
CISA Zero Trust Maturity Model 2.0: https://www.cisa.gov/sites/default/files/2023-04/zero_trust_maturity_model_v2_508.pdf