NISTAML.031 · Model Extraction

Explore 9 source-supported AI security cases for NISTAML.031 · Model Extraction in NIST Adversarial ML (2025), with evidence, scope, and defensive analysis.

Risk category

An attacker extracts information about a model's architecture or parameters, commonly by submitting carefully chosen queries to a machine-learning service. Methods may also exploit side-channel information, and the result may be a functionally similar model rather than an exact copy. Learning these details can expose proprietary assets and help the attacker design stronger attacks against the original model.

Official framework source ↗ · Open interactive explorer ↗

Incident · · Practical overlap

Industrial-Scale Distillation Across U.S. Frontier AI Services

NSA, CISA and FBI attributed coordinated distillation campaigns to six China-based AI companies. Their September 2026 advisory describes fraudulent accounts, bulk subscriptions, gray-market proxies, cross-provider routing, request-metadata removal and hidden-reasoning prompts, followed by response collection and quality checks. Distillation trains a model from another model’s outputs; the advisory’s attribution and capability-assessment claims are attributed to the agencies, not presented as independently reproduced measurements or model-weight theft.

Why it relates: Response queries and downstream training in the campaigns attributed by NSA, CISA and FBI aim to recreate selected model behavior. The demonstrated target is capability replication. Architecture inference, parameter recovery and exact model equality are not claimed.

the campaigns attributed by NSA, CISA and FBI; capability distillation is distinguished from stealing original weights. Attribution and capability outcomes remain agency assessments. Hidden-reasoning prompts are attempts, not proof that the returned text was an internal trace. No autonomous agent deployment is inferred.

SecureFlow · Read the full case ↗

Incident · · Practical overlap

Model Distillation Campaigns Target Google AI Services

Google reported campaigns that used compromised credentials, fraudulent accounts and proxies to collect model outputs for distillation: training another model from a provider’s responses. A February 2026 reasoning-focused campaign exceeded 100,000 prompts; later campaigns included some above 100 million. Those are campaign query volumes, not model weights stolen or a measured replication success rate. Google says it disrupted accounts and reduced the usefulness of collected outputs.

Why it relates: Response queries and downstream training in Google’s compromised-account and fraudulent-account campaigns aim to recreate selected model behavior. The demonstrated target is capability replication. Architecture inference, parameter recovery and exact model equality are not claimed.

Google’s compromised-account and fraudulent-account campaigns; capability distillation is distinguished from stealing original weights. The prompt totals are collection volumes, not measured replication success or model weights. No autonomous agent deployment or protected-reasoning recovery is established.

SecureFlow · Read the full case ↗

Incident · · Practical overlap

Model Distillation Campaigns Targeting Anthropic Claude

Anthropic attributed more than 16 million exchanges through roughly 24,000 fraudulent accounts to DeepSeek, Moonshot AI and MiniMax. The flow distinguishes proxy access, authenticated accounts, distributed routing, inference-response collection and downstream model training. Provider-side hunting and verified limits act before outputs leave; potential harms from models lacking safeguards remain Anthropic’s risk assessment, not observed downstream victim counts.

Why it relates: Response queries and downstream training in Anthropic’s DeepSeek, Moonshot AI and MiniMax campaigns aim to recreate selected model behavior. The demonstrated target is capability replication. Architecture inference, parameter recovery and exact model equality are not claimed.

Anthropic’s DeepSeek, Moonshot AI and MiniMax campaigns; capability distillation is distinguished from stealing original weights. The 24,000-account and 16-million-exchange figures do not quantify recovered weights or observed downstream harms. Potential harms remain Anthropic’s assessment; an agent deployment is not established.

SecureFlow · Read the full case ↗

Validated Research · · Practical overlap

Confusing Antimalware Neural Networks

Cloud storage and computations have become popular platforms for deploying ML malware detectors. In such cases, the features for models are built on users' systems and then sent to cybersecurity company servers. The Kaspersky ML research team explored this gray-box scenario and showed that feature knowledge is enough for an adversarial attack on ML models. They attacked one of Kaspersky's antimalware ML models without white-box access to it and successfully evaded detection for most of the adversarially modified malware files.

Why it relates: Adversarial feature knowledge enables gray-box evasion of an antimalware model; a pure black-box threat model is not established. The researchers scan a labeled dataset with the cloud detector, train a proxy on its predictions, reverse-engineer the local feature extractor and transfer adversarial files back to the target. The cloud model itself remains a black box. The complete system is gray-box because the local feature extractor is inspected. Only the cloud-model query boundary and prediction-trained surrogate are shared; original weights and exact architecture are not recovered.

Cloud storage and computations have become popular platforms for deploying ML malware detectors. In such cases, the features for models are built on users' systems and then sent to cybersecurity company servers. The Kaspersky ML research team explored this gray-box scenario and showed that feature knowledge is enough for an adversarial attack on ML models. They attacked one of Kaspersky's antimalware ML models without white-box access to it and successfully evaded detection for most of the adversarially modified malware files.

SecureFlow · Read the full case ↗

Validated Research ·

ProofPoint Evasion

Proof Pudding (CVE-2019-20634) is a code repository that describes how ML researchers evaded ProofPoint's email protection system by first building a copy-cat email protection ML model, and using the insights to bypass the live system. More specifically, the insights allowed researchers to craft malicious emails that received preferable scores, going undetected by the system. Each word in an email is scored numerically based on multiple variables and if the overall score of the email is too low, ProofPoint will output an error, labeling it as SPAM.

Why it relates: Proof Pudding (CVE-2019-20634) is a code repository that describes how ML researchers evaded ProofPoint's email protection system by first building a copy-cat email protection ML model, and using the insights to bypass the live system. The relevant category is Model Extraction: An attacker extracts information about a model's architecture or parameters, commonly by submitting carefully chosen queries to a machine-learning service.

Proof Pudding (CVE-2019-20634) is a code repository that describes how ML researchers evaded ProofPoint's email protection system by first building a copy-cat email protection ML model, and using the insights to bypass the live system. More specifically, the insights allowed researchers to craft malicious emails that received preferable scores, going undetected by the system. Each word in an email is scored numerically based on multiple variables and if the overall score of the email is too low, ProofPoint will output an error, labeling it as SPAM.

SecureFlow · Read the full case ↗

Validated Research ·

Attack on Machine Translation Services

Machine translation services (such as Google Translate, Bing Translator, and Systran Translate) provide public-facing UIs and APIs. A research group at UC Berkeley utilized these public endpoints to create a replicated model with near-production state-of-the-art translation quality. Beyond demonstrating that IP can be functionally stolen from a black-box system, they used the replicated model to successfully transfer adversarial examples to the real production services. These adversarial inputs successfully cause targeted word flips, vulgar outputs, and dropped sentences on Google Translate and Systran Translate websites.

Why it relates: Machine translation services (such as Google Translate, Bing Translator, and Systran Translate) provide public-facing UIs and APIs. The relevant category is Model Extraction: An attacker extracts information about a model's architecture or parameters, commonly by submitting carefully chosen queries to a machine-learning service.

Machine translation services (such as Google Translate, Bing Translator, and Systran Translate) provide public-facing UIs and APIs. A research group at UC Berkeley utilized these public endpoints to create a replicated model with near-production state-of-the-art translation quality. Beyond demonstrating that IP can be functionally stolen from a black-box system, they used the replicated model to successfully transfer adversarial examples to the real production services. These adversarial inputs successfully cause targeted word flips, vulgar outputs, and dropped sentences on Google Translate and Systran Translate websites.

SecureFlow · Read the full case ↗

Validated Research · · Practical overlap

GPUBreach: An AI GPU Boundary Can Fail Below the Model

GPUBreach researchers used unprivileged CUDA code on an NVIDIA RTX A6000 to corrupt GPU page tables through Rowhammer, a repeated-memory-access technique that induces bit flips. Their controlled demonstrations include cross-process GPU access, LLM weight leakage and a separate driver path to CPU root with IOMMU enabled. The relevant AI defense is removing untrusted GPU co-residency, not changing a model prompt.

Why it relates: Unauthorized parameter access exposes the confidential model implementation. This is direct GPU-memory theft, not API-based functional cloning or a demonstrated copy of every model parameter.

Controlled GPU page-table corruption and distinct CPU-driver escalation; model-weight theft is not training-data inference.

AIA · Read the full case ↗

Validated Research ·

Google Photos AI Model Extraction

Skyld researchers analyzed the Google Photos Android application and recovered TensorFlow Lite models used by AI-powered photo editing and image analysis features. The researchers found models stored unencrypted in the application's assets, embedded in the native library, and encrypted on disk. They used static reverse engineering to locate TFLite artifacts and dynamic instrumentation with Frida to capture encrypted models after runtime decryption. The recovered models provided white-box access to proprietary Google Photos model artifacts.

Why it relates: Static reverse engineering and runtime decryption interception expose model architecture and parameters.

Static APK/native-library extraction and runtime interception recover model files. The demonstrated result is model access, with later evasion only potential.

SecureFlow · Read the full case ↗

Exercise · · Practical overlap

Microsoft Azure Service Disruption

The Microsoft AI Red Team performed a red team exercise on an internal Azure service with the intention of disrupting its service. This operation had a combination of traditional ATT&CK enterprise techniques such as finding valid account, and exfiltrating data -- all interleaved with adversarial ML specific steps such as offline and online evasion examples.

Why it relates: The Azure-service exercise targets availability across separate conventional and ML steps; it is not an observed malicious compromise. The team exfiltrates the target model file before offline attack development, gaining access to the model asset rather than learning only its predictions. The shared result is loss of model confidentiality through direct artifact theft. The source does not demonstrate query-based parameter reconstruction or functional distillation.

The Microsoft AI Red Team performed a red team exercise on an internal Azure service with the intention of disrupting its service. This operation had a combination of traditional ATT&CK enterprise techniques such as finding valid account, and exfiltrating data -- all interleaved with adversarial ML specific steps such as offline and online evasion examples.

SecureFlow · Read the full case ↗