What type of study is this?

This is a Systematic Review study.

October 3, 2025Open Access

Attack and defense techniques in large language models: A survey and new perspectives

View Full Paper

Ask AI

Bookmark

Authors

This systematic survey explores attack methods and defense strategies in large language models, indicating challenges in security and usability.

The survey reveals vulnerabilities in large language models and highlights critical security challenges.
Attacks are classified into adversarial prompt attacks and model theft, outlining their mechanisms and implications.
Defense strategies include prevention-based and detection-based methods, emphasizing the importance of adaptability.
Challenges in implementing defenses are discussed, including the need for explainable security techniques and standardized evaluations.

Liao et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

This systematic survey explores attack methods and defense strategies in large language models, indicating challenges in security and usability.