Scholarlink

Abstract <p>Modern large language models possess impressive capabilities but remain vulnerable to various attacks capable of manipulating their responses, causing confidential data leaks, or bypassing restrictions. The main focus is on analyzing “prompt injection” attacks, which allow circumventing model limitations, extracting hidden data, or forcing the model to follow malicious instructions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From Exploitation to Protection: Analysis of Attacks on Large Language Models

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • S. V. Bezzateev,
  • I. S. Velichko

摘要

Abstract

Modern large language models possess impressive capabilities but remain vulnerable to various attacks capable of manipulating their responses, causing confidential data leaks, or bypassing restrictions. The main focus is on analyzing “prompt injection” attacks, which allow circumventing model limitations, extracting hidden data, or forcing the model to follow malicious instructions.