# What is Adversarial Machine Learning?

> Adversarial machine learning is the field concerned with attacks that exploit machine-learning systems and with methods for understanding and mitigating those attacks.

- Canonical URL: https://yellowcube.eu/glossary/adversarial-machine-learning/
- Publisher: Yellow Cube
- Language: en
- Contact: hello@yellowcube.eu

## Content

It examines how an adversary can influence data, models, inputs, outputs or surrounding workflows to violate confidentiality, integrity, availability or acceptable-use objectives. It is a body of attack and defense knowledge, not the name of one specific exploit.

NIST’s current taxonomy organizes attacks using dimensions such as the lifecycle stage, the adversary’s goal, objective, capability and knowledge. Relevant families include evasion at inference time, poisoning during training, privacy attacks and abuse of generative systems through prompting or misuse. Real systems also inherit conventional weaknesses in identities, APIs, software, cloud services and supply chains, so an adversarial-ML assessment should connect model behavior to the complete application.

### Key points

- **Attack stage:** Determine whether the adversary can affect data collection, training, model distribution, deployment, inference inputs, retrieval, feedback or connected actions.
- **Adversary conditions:** State what access, knowledge, control, resources and query opportunities an attack assumes; a result under full model access may not transfer to a remote service.
- **Security objectives:** Evaluate availability failures, untargeted or targeted integrity violations, privacy loss, model theft and misuse against consequences meaningful to the deployed context.
- **Defensive practice:** Combine model-level evaluation with provenance, access control, secure engineering, monitoring, incident response and constraints on downstream decisions or actions.
- **Important limitation:** Robustness against a known benchmark or attack method is not general security. Attackers adapt, research assumptions vary, and controls that improve one property may leave application or human trust boundaries exposed.

### Related terms

[AI security](<https://yellowcube.eu/glossary/ai-security/>) · [Data poisoning](<https://yellowcube.eu/glossary/data-poisoning/>) · [Model poisoning](<https://yellowcube.eu/glossary/model-poisoning/>) · [Prompt injection](<https://yellowcube.eu/glossary/prompt-injection/>) · [AI red teaming](<https://yellowcube.eu/glossary/ai-red-teaming/>)

### Sources

[NIST AI 100-2 E2025, Adversarial Machine Learning Taxonomy](https://csrc.nist.gov/pubs/ai/100/2/e2025/final) · [NIST AI Security and Resilience](https://www.nist.gov/artificial-intelligence/ai-research-security-and-resilience) · [MITRE ATLAS](https://atlas.mitre.org/)

## Attribution and scope

This Markdown representation is generated from the same approved content records as the canonical HTML page. Cite the canonical URL above when referencing this material. Product and service descriptions are informational; confirm project-specific requirements with Yellow Cube.