paperJune 2026Unreviewed

LLM Security Defense Using Unsupervised Learning and Deep Reinforcement Learning

Leilei Wang, Hongying Li

2026 IEEE 3rd International Conference on Computer Vision and Deep Learning (DLCV)

Abstract

With the widespread deployment of large language models (LLMs) in intelligent systems, security threats such as prompt injection, jailbreaking, data poisoning, and hidden backdoor attacks have become increasingly severe. Traditional rule-based filtering and static detection methods are unable to capture implicit semantic attacks and hidden-space anomalies, and lack adaptability to evolving adversarial behaviors. To address these challenges,, unsupervised feature learning, deep reinforcement lear

Framework mappings

OWASP Top 10 for LLM Applications

LLM01Prompt Injection
LLM04Data and Model Poisoning

MITRE ATLAS

AML.T0020Poison Training Data
AML.T0051LLM Prompt Injection
AML.T0054LLM Jailbreak

Suggested from the entry's categories.

Cite

@inproceedings{wang2026llm,
  title = {{LLM Security Defense Using Unsupervised Learning and Deep Reinforcement Learning}},
  author = {Leilei Wang and Hongying Li},
  year = {2026},
  month = jun,
  booktitle = {2026 IEEE 3rd International Conference on Computer Vision and Deep Learning (DLCV)},
  doi = {10.1109/DLCV69906.2026.11635183},
  url = {https://www.semanticscholar.org/paper/26c28b6a8e743b6d6f8ff199fb2217b5d7fa9f34}
}