122 lines
9.9 KiB
JSON
122 lines
9.9 KiB
JSON
[
|
||
{
|
||
"title": "以大模型为目标的威胁攻击与安全思考\n",
|
||
"link": "https://paper.seebug.org/3259/",
|
||
"description": "作者:启明星辰ADLab\n原文链接:https://mp.weixin.qq.com/s/YL37d_PjiFJkarEjcrTdiA\n一、概 述\n自2023年以来,以ChatGPT为代表的大语言模型(LLM)向人们展现出了人工智能所具有的无限潜力,为各行业带来了全新的发展模式和生产力提升。以大语言模型为基础的智能平台及系统也逐渐成为国家和企业的一种重要的基础设施,面对这样一种全新的基础设施...\n",
|
||
"pubDate": "Mon, 23 Dec 2024 08:21:00 +0000",
|
||
"guid": "https://paper.seebug.org/3259/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "趋利避害的大模型行为:伪对齐\n",
|
||
"link": "https://paper.seebug.org/3258/",
|
||
"description": "作者:Ryan Greenblatt, Carson Denison等\n译者:知道创宇404实验室翻译组\n原文链接:https://arxiv.org/abs/2412.14093v1\n摘要\n我们在此呈现了一项关于大型语言模型在训练过程中进行“对齐伪装”行为的演示:该模型有选择地遵从其训练目标,以防止其行为在训练之外被改变。具体来说,我们首先向Claude 3 Opus模型提供了一个系统提示...\n",
|
||
"pubDate": "Mon, 23 Dec 2024 03:06:00 +0000",
|
||
"guid": "https://paper.seebug.org/3258/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "找出披着羊皮的狼:通过文本摘要技术破解对抗性文本到图像提示\n",
|
||
"link": "https://paper.seebug.org/3257/",
|
||
"description": "作者:Portia Cooper, Harshita Narnoli, Mihai Surdeanu\n译者:知道创宇404实验室翻译组\n原文链接:https://arxiv.org/pdf/2412.12212\n摘要\n文本到图像模型常常遭受一种称为“分而治之攻击”(DACA)的逐步攻击,该攻击通过大型语言模型将敏感文本包装在看似无害的叙述中,以隐藏不当内容。为了有效抵御这种逐步DACA攻击,...\n",
|
||
"pubDate": "Fri, 20 Dec 2024 07:43:00 +0000",
|
||
"guid": "https://paper.seebug.org/3257/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "探索 AI 驱动的网络安全框架:深度学习技术、GPU 支持和未来增强\n",
|
||
"link": "https://paper.seebug.org/3255/",
|
||
"description": "作者:Tobias Becher, Simon Torka\n译者:知道创宇404实验室翻译组\n原文链接:https://arxiv.org/pdf/2412.12648\n摘要\n传统的基于规则的网络安全系统在防御已知恶意软件方面表现出色,但面对新型威胁,它们却显得力不从心。为应对这一挑战,新一代网络安全系统开始融合人工智能技术,尤其是深度学习算法,以提升其识别事件、分析警报和响应安全事件的能力...\n",
|
||
"pubDate": "Fri, 20 Dec 2024 01:14:00 +0000",
|
||
"guid": "https://paper.seebug.org/3255/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "Apache Struts2 文件上传逻辑绕过(CVE-2024-53677)(S2-067)\n",
|
||
"link": "https://paper.seebug.org/3256/",
|
||
"description": "作者:y4tacker\n原文链接:https://y4tacker.github.io/2024/12/16/year/2024/12/Apache-Struts2-文件上传逻辑绕过-CVE-2024-53677-S2-067/\n前言\nApache官方公告又更新了一个Struts2的漏洞,考虑到很久没有发无密码的博客了,再加上漏洞的影响并不严重,因此公开分享利用的思路。\n分析\n影响版本\nSt...\n",
|
||
"pubDate": "Thu, 19 Dec 2024 08:37:00 +0000",
|
||
"guid": "https://paper.seebug.org/3256/",
|
||
"category": "漏洞分析"
|
||
},
|
||
{
|
||
"title": "越狱破解马斯克最新AI-Grok2揭秘:特朗普赢得大选背后,AI的推波助澜\n",
|
||
"link": "https://paper.seebug.org/3254/",
|
||
"description": "作者:洺熙(米斯特Ai安全组核心成员)\n原文链接:https://mp.weixin.qq.com/s/thSJwKA7RbEueQ7iYsThFQ\n序言\n你可曾想过,马斯克的Ai Grok2大脑里究竟藏着什么秘密?\n本文将带你潜入X平台最新AI模型Grok2的“大脑”,通过对其核心指令——系统Prompt的深度解码,揭开马斯克打造的这款AI的神秘面纱\n我们将一起探索:\n\n从越狱谷歌Gemi...\n",
|
||
"pubDate": "Wed, 18 Dec 2024 06:02:00 +0000",
|
||
"guid": "https://paper.seebug.org/3254/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "AI 作为新型黑客:开发进攻性安全代理\n",
|
||
"link": "https://paper.seebug.org/3253/",
|
||
"description": "作者:Leroy Jacob Valencia\n译者:知道创宇404实验室翻译组\n原文链接:https://arxiv.org/pdf/2406.07561\n摘要\n在网络安全这一宏大领域内,由防御性措施向主动防御的转变对于守护数字基础设施的安全至关重要。本文深入探讨了AI在主动防御网络安全中的角色,尤其是通过研发一个名为ReaperAI的独立AI代理,该代理被设计来模拟和实施网络攻击。借助于...\n",
|
||
"pubDate": "Fri, 13 Dec 2024 07:50:00 +0000",
|
||
"guid": "https://paper.seebug.org/3253/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "Qwen2.5-Coder 技术报告\n",
|
||
"link": "https://paper.seebug.org/3252/",
|
||
"description": "公众号:简单的机器学习\n原文链接:https://mp.weixin.qq.com/s/EiV7x403sVqVcABo_qd2kg\n引言\nQwen2.5-Coder 系列是阿里巴巴团队推出的一款重要的代码生成模型,相比其前代 CodeQwen1.5,该系列在多个方面进行了显著的升级。Qwen2.5-Coder 系列包括两个模型:Qwen2.5-Coder-1.5B 和 Qwen2.5-Co...\n",
|
||
"pubDate": "Wed, 11 Dec 2024 16:46:00 +0000",
|
||
"guid": "https://paper.seebug.org/3252/",
|
||
"category": "经验心得"
|
||
},
|
||
{
|
||
"title": "Qwen2 源码阅读——核心代码跟读\n",
|
||
"link": "https://paper.seebug.org/3251/",
|
||
"description": "公众号:简单的机器学习\n原文链接:https://mp.weixin.qq.com/s/PVSPNfv0I8_cxgPTmOes5w\n我们继续使用上一节使用的样例文本:\n["你好啊", "简单的机器学习是为了让机器学习变得更简单而存在的"]\n\n\n这个样例产生的tokens结果为:\n{'input_ids': tensor([[108386, 10392...\n",
|
||
"pubDate": "Wed, 11 Dec 2024 16:45:00 +0000",
|
||
"guid": "https://paper.seebug.org/3251/",
|
||
"category": "经验心得"
|
||
},
|
||
{
|
||
"title": "Qwen2 源码阅读——环境准备和说明\n",
|
||
"link": "https://paper.seebug.org/3250/",
|
||
"description": "公众号:简单的机器学习\n原文链接:https://mp.weixin.qq.com/s/kxrc50ZumITVaTE1wtinlg\n下面的源码内容来自transformers代码库中:transformers-4.45.2/src/transformers/models/qwen2/modeling_qwen2.py。\n实验准备\n首先我们下载一些Qwen2需要的配置数据。下载地址:http...\n",
|
||
"pubDate": "Wed, 11 Dec 2024 16:24:00 +0000",
|
||
"guid": "https://paper.seebug.org/3250/",
|
||
"category": "经验心得"
|
||
},
|
||
{
|
||
"title": "揭秘暗黑系网络服务运营商——防弹主机网络\n",
|
||
"link": "https://paper.seebug.org/3248/",
|
||
"description": "作者:知道创宇404实验室\n日期:2024年12月9日\nEnglish version: https://paper.seebug.org/3249/\n一.摘要\n防弹主机(Bulletproof hosting)托管服务为网络犯罪活动提供基础设施,允许犯罪分子规避法律约束,经常被用于恶意软件、黑客攻击、欺诈网站、垃圾邮件等。防弹主机网络,堪称暗黑系网络服务运营商,是专为网络犯罪活动提供恶意基...\n",
|
||
"pubDate": "Wed, 11 Dec 2024 08:06:00 +0000",
|
||
"guid": "https://paper.seebug.org/3248/",
|
||
"category": "专题报告"
|
||
},
|
||
{
|
||
"title": "Unveiling Dark Internet Service Providers: Bulletproof Hosting\n",
|
||
"link": "https://paper.seebug.org/3249/",
|
||
"description": "Author: Knownsec 404 team\nDate: Dec 9, 2024\n中文版:https://paper.seebug.org/3248/\n1. Abstract\nBulletproof hosting services provide the infrastructure for cybercriminal activities, enabling criminals t...\n",
|
||
"pubDate": "Mon, 09 Dec 2024 04:14:00 +0000",
|
||
"guid": "https://paper.seebug.org/3249/",
|
||
"category": "404 English Paper"
|
||
},
|
||
{
|
||
"title": "基于 InternLM 和 LangChain 搭建私人知识库\n",
|
||
"link": "https://paper.seebug.org/3247/",
|
||
"description": "作者:Jinzhong Xu\n原文链接:https://xujinzh.github.io/2024/01/08/ai-Interlm-langchain-RAG/index.html\n本篇介绍基于 InternLM 和 LangChain 搭建私人知识库。\npython!conda create --name internlm_langchain --clone=/root/share/c...\n",
|
||
"pubDate": "Thu, 05 Dec 2024 11:34:00 +0000",
|
||
"guid": "https://paper.seebug.org/3247/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "书生·浦语大模型使用\n",
|
||
"link": "https://paper.seebug.org/3246/",
|
||
"description": "作者:Jinzhong Xu\n原文链接:https://xujinzh.github.io/2024/01/06/ai-internlm-useage/\n本篇介绍书生·浦语大模型的使用,包括智能对话、智能体工具调用和图文理解创作等。\n环境配置\npython!conda create --name internlm-chat --clone=/root/share/conda_envs/int...\n",
|
||
"pubDate": "Thu, 05 Dec 2024 10:27:00 +0000",
|
||
"guid": "https://paper.seebug.org/3246/",
|
||
"category": "AI安全"
|
||
},
|
||
{
|
||
"title": "Uncontrolled PCDN: Observation and Case Analysis of the Current State of PCDN Technology\n",
|
||
"link": "https://paper.seebug.org/3245/",
|
||
"description": "Author: Knownsec 404 Active Defense Team\nDate: November 22, 2024\n中文版:https://paper.seebug.org/3242\nBackground Introduction\nIn October 2024, Knownsec 404 Active Defense Team detected abnormal traffi...\n",
|
||
"pubDate": "Mon, 02 Dec 2024 06:13:00 +0000",
|
||
"guid": "https://paper.seebug.org/3245/",
|
||
"category": "情报分析"
|
||
}
|
||
] |