Skip to main content

AI/ML / Gen AI / Controls / DEV

Model Input Filtering and Sanitisation

CCC.GenAI.CN01 · MachineLearning

Inspect and validate input before it is passed to a GenAI model in order to filter or sanitise adversarial queries and prevent sensitive data leakage.

Related Capabilities

IDTitleDescription
CCC.Core.CP14API AccessThe service exposes a port enabling external actors to interact programmatically with the service and its resources using HTTP protocol methods such as GET, POST, PUT, and DELETE.
CCC.GenAI.CP15Text-Based PromptsAbility to input prompts in plain text.
CCC.GenAI.CP16Structured PromptsAbility to provide structured input such as JSON as prompts.
CCC.GenAI.CP17Contextual PromptsAbility to provide context or background information within the prompt to guide the response.
CCC.GenAI.CP18Interactive PromptsAbility to use conversational prompts to create interactive dialogues.
CCC.GenAI.CP19Image-Based PromptsAbility to input an image as a prompt to generate a response.
CCC.GenAI.CP20Custom Template PromptsAbility to define custom templates or structures for prompts to standardize interactions with the models.
CCC.GenAI.CP21Generate ContentAbility to generate a response given a foundation model, parameter values, and a prompt.
CCC.GenAI.CP24Content ModerationEnsure the service detects and filters abusive, harmful, and sensitive information to ensure responsible and safe use of the service.
CCC.Core.CP02Encryption at Rest Enabled by DefaultThe service automatically encrypts all data using industry-standard cryptographic protocols prior to being written to a storage medium.
CCC.Core.CP06Access ControlThe service automatically enforces user configurations to restrict or allow access to a specific component or a child resource based on factors such as user identities, roles, groups, or attributes.
CCC.GenAI.CP22Data ControlEnsures prompts, model outputs, embeddings, and training data fed by customers are not used to train foundation models.

Related Threats

IDTitleDescription
CCC.GenAI.TH01Prompt InjectionPrompt injection may occur when crafted input is used to manipulate the GenAI model's behaviour, resulting in the generation of harmful or unintended outputs. Prompt injection can be either direct (performed via direct interaction with the model) or indirect (performed via external sources ingested by the model). Both text-based and multi-modal prompt injection is possible.
CCC.GenAI.TH03Sensitive Information DisclosureSensitive data can be memorised by the model from user interaction or training and may then be leaked to unintended and unauthorised parties by querying the model, for example through crafted prompts.

Assessment Requirements

IDTextApplicability
CCC.GenAI.CN01.AR01Untrusted input such as user queries, RAG data or tool output MUST be validated before it is passed to a GenAI model.tlp-clear, tlp-green, tlp-amber, tlp-red
CCC.GenAI.CN01.AR02If malicious patterns such as prompt injection or sensitive data are detected during input validation, the input MUST be blocked or sanitised.tlp-clear, tlp-green, tlp-amber, tlp-red

Guideline Mappings

FrameworkIDRemarks
FINOS-AIGFAIR-PREV-003User/App/Model Firewalling/Filtering
FINOS-AIGFAIR-PREV-017AI Firewall Implementation and Management
FINOS-AIGFAIR-PREV-002Data Filtering From External Knowledge Bases
FINOS-AIGFAIR-DET-001AI Data Leakage Prevention and Detection
SAIFInput Validation and Sanitization
MITRE-ATLASAML.M0020Generative AI Guardrails
MITRE-ATLASAML.M0021Generative AI Guidelines
MITRE-ATLASAML.M0015Adversarial Input Detection