Aggregator

在litellm-1.82.8、litellm-1.82.7包的proxy_server.py代码中存在多段不同的b64_payload代码载荷，推测其是历史版本载荷。

Five Practical Use Cases on How AI Is Transforming SOCs for Threat Mitigation
AI is reshaping cybersecurity on both sides of the battlefield. While attackers use it to scale threats, defenders are using AI to reduce noise, accelerate investigations and improve response. This blog provides five real-world SOC use cases for mitigating threats.

Expect Fallout After Remote Access Trojan Added to Popular JavaScript NPM Package
A supply-chain attack backdoored versions of Axios, a popular JavaScript library that's present in many different software packages, to distribute a cross-platform, remote access Trojan. Identifying the full fallout from the attack could take some time, experts warned.

Chrome Zero-Day CVE-2026-5281: A Use-After-Free in Dawn's WebGPU Layer

These are the top threats you should know about this week.

结合代码浅分析DigitalOcean Droplet Agent 命令注入预授权 RCE 漏洞(CVE-2026-24516)

本文记录一次完整的内网渗透实战过程，从外网打点到域控接管，涵盖 WebShell 利用、OA 漏洞利用、内网横向移动、域提权等关键技术。

TexSaw CTF 2026 Writeup

本文分析了SUCTF题目SU_jdbc-master的完整攻击链。利用Unicode字符ſ转大写为S的特性，绕过仅做小写匹配的Spring MVC鉴权拦截器；随后通过Jackson反序列化覆盖默认JDBC驱动为存在漏洞的Kingbase8，结合临时文件写入与文件描述符爆破，触发恶意Spring XML配置加载，最终实现命令执行回显。

第十九届ciscn&长城杯半决赛 isw部分wp分享

1.Illusion[hook]表面上看是 RC4 其实被hook到了AES上 这题表面主逻辑是 RC4，但真正的 flag 在 MessageBoxA 钩子里的 AES 分支里分析main函数前缀固定为 xmctf{总长度固定为 25} 在最后一位中间 18 字节会被取出来参与校验sub_140001440 很像标准 RC4：初始化 256 字节 S 盒用密钥 nev_gona_give_

当 AI Agent 拥有了读写文件、执行命令、访问网络的能力，如何用平台自身的扩展机制构建安全防线？

仿真内网、内网渗透、横向移动、权限维持、多层代理、evasion

A vulnerability, which was classified as problematic, was found in Free5GC 4.2.0. The affected element is an unknown function of the component aper. Such manipulation leads to type confusion. This vulnerability is traded as CVE-2026-5360. The attack may be launched remotely. Furthermore, there is an exploit available. It is best practice to apply a patch to resolve this issue.

随着HTTPS加密通信的普及，传统基于明文特征的流量检测方法逐渐失效，恶意软件也越来越多地利用TLS协议隐藏其通信行为，给安全检测带来了新的挑战。针对这一问题，JA3指纹技术通过对TLS ClientHello消息中的关键字段进行特征提取与哈希计算，实现了在加密流量环境下对客户端行为的有效识别。

我们可以进一步利用CVE-2026-26023漏洞，在受害者查看历史聊天记录时触发恶意脚本，进而借助受害者浏览器的已授权上下文，调用Dify后端API，最终实现从中危XSS到窃取工作空间权限的危害提升

A vulnerability, which was classified as critical, has been found in DedeCMS 5.7.118. Impacted is an unknown function of the component Module Handler. This manipulation of the argument setup tag causes unrestricted upload. This vulnerability appears as CVE-2026-30643. The attack may be initiated remotely. There is no available exploit.

A vulnerability classified as problematic was found in streetwriters notesnook up to 3.3.16. This issue affects some unknown processing. The manipulation results in cross site scripting. This vulnerability is reported as CVE-2026-33978. The attack can be launched remotely. No exploit exists. Upgrading the affected component is advised.

A vulnerability classified as critical has been found in Sinaptik AI pandas-ai 3.0.0. This vulnerability affects the function pandasai.agent.base._execute_sql_query. The manipulation leads to sql injection. This vulnerability is documented as CVE-2026-30273. The attack can be initiated remotely. There is not any exploit available.