AI projects face security challenges resulting from difficulties in ensuring model integrity and reliability. The Sleepy Pickle and HuggingFace models' silent backdoors are notable examples of such model security loopholes.
Understanding the AI security challenge
AI project components comprise data, models, and code. Modifying the data or code often allows for the change of the model component without a trace. Thorough documentation of the creation, operation, and lifecycle of your model’s dependencies, processes, metadata, artifacts, tools, libraries, etc. enables you to protect your AI system from common model breaches like data or model poisoning, model evasion, confidentiality attacks, and model flaws.
Why model attestation and SBOMs matter for your AI project
Model attestation, similar to a traditional software attestation security strategy, is a practice that enables you to establish a verifiable security supply chain for all system components. It is a strategy that verifies the integrity, authenticity, and lifecycle of your model's data, code, and artifacts and their relationship at the different stages of the development lifecycle.
Software Bill of Materials (SBOMs) or AI Bill of Materials (AIBOM), in the context of AI projects, is an inventory report you can use for model attestation. It provides an inventory of the components and the dependencies used in building a model. This inventory report should typically include data, model artifacts, dependencies, and their relationships.
To know more about the topic and the tools that will help in tackling these problems of AI security, check out this full blog by Jozu.