# What is Adversarial Attack? Definition & Technical Mechanics

> **Source:** [SPIDITS AI AI Glossary](https://spidits.com/ai-glossary)  
> **Category:** Model Limitations  
> **Canonical Citation:** [https://spidits.com/ai-glossary/adversarial-attack](https://spidits.com/ai-glossary/adversarial-attack)

## Definition
An Adversarial Attack is a technique that feeds a machine learning model intentionally designed inputs (adversarial examples) to cause it to make a mistake, fail, or hallucinate. In image models, this often involves introducing imperceptible pixel noise that completely alters the classification.

## Architectural Mechanics & Deep Dive
Adversarial attacks represent a major security risk for machine learning models, where inputs are subtly modified to deliberately fool a model into making incorrect predictions. In computer vision, this can involve injecting imperceptible pixel noise into an image, causing a classifier to identify a stop sign as a speed limit sign. In Large Language Models (LLMs), adversarial attacks manifest as prompt injections or jailbreaks, where users craft specific input prompts to bypass safety filters and elicit restricted information.

## Industry Applications & Real-World Use Cases
Security vulnerability audits, defense hardening, and robustness testing

## Frequently Asked Questions

### How do you defend against adversarial attacks?
By performing adversarial training, where adversarial examples are generated and included directly in the training dataset to build model robustness.

### What is a jailbreak in LLMs?
A jailbreak is a text-based adversarial attack where a user structures prompts to bypass the safety alignment filters of a Large Language Model.

## Related Terminology
- [hallucination](https://spidits.com/ai-glossary/hallucination)
- [llm](https://spidits.com/ai-glossary/llm)
- [guardrails](https://spidits.com/ai-glossary/guardrails)

---
*From the SPIDITS AI Knowledge Base (336+ terms indexed). Explore full technical definitions: [https://spidits.com/ai-glossary](https://spidits.com/ai-glossary)*