WTF????????
*sshole is out
if you run sshd on alpine, openwrt or any musl box with password auth on: update now
official writeup: sshhole.com
5
1,583
Introducing ZeroMDASH with @bestswngs : open-source, agent-native vulnerability research.
Specs, target/similar-software CVEs, writeups, patch analysis & new features → hypotheses → investigator/critic verification → new hypotheses → ... → attack chains → reports.
LangGraph-based. Cases: KVM guest eret2host + Milvus heap overflow. cossack9989.github.io/slides…
github.com/JANUS-SEC/ZeroMDA…
in-tree kvm 默认开启 nested virtualization,但是如果有专职虚拟化团队,往往会发生以下两件事
1. 团队维护 out-of-tree kvm dkms,nv 会因为在 L1 Guest 没太大业务价值而被关闭
2. 如果一定要开 L2 Guest(比如某些 region 实在开不出裸金属),优先在 L1 Guest 安装 out-of-tree pvm dkms
We’ve confirmed a KVM 0day through our Vercel Sandbox bounty program. Affecting the industry’s gold standard solution for Linux virtualization.
2026 is wild! Thankful to Paulos and other researchers helping us make the most secure sandbox for agents. Full writeup coming.
11
2,341
Agent Infra 安全领域终于有了点曝光度~
不过话又说回来,基本每家 Agent 大厂都有个类似 #NVIDIA OpenShell 的 Agent Infra 吧?各种 Sandbox 混跑 + Host 侧导流 Sandbox 内的流量到一个支持 4 层和 7 层访问控制的 Agent Gateway
BF4 Sentry 估计对应 BF3 的 Argus,原理类似 DMA 外挂扫内存,有想象空间
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.
Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come.
But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility.
This is bigger than a single product. It's the beginning of an open ecosystem to build the trust layer for safe agent systems.
Together, we are building the foundation of the AI economy.
Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. nvda.ws/4hcoq7m
摆烂了,别卷挖洞了
两个月前写了个微软MDASH翻版,用GLM-5.2(代码审计) + GPT-5.5(Context采集)组了个巨大的workflow,扫某固件源码,发现了些high、critical的漏洞,成本¥1k
今天试了下dsh minimal mode + deepseek-v4-pro-0813,发现了几乎完全一样的漏洞集合(只缺了2个),成本4¥
最近想到,agent infra的vm超分时,一机一client的HIDS会占用与vm数量成正比的资源,所以是不是该重拾vmi把HIDS卸载到宿主了?
恰好今天在看蚂蚁的pvm,有guest->switcher->guest的路径,于是蹬gpt给pvm搓了vmi,不过纯demo,只跑了stress-ng和openssl-speed(
github.com/virt-pvm/linux/co…
3
1,257
发表一些暴论:
传统安全研究员在3T模型出现后、AGI实现前,除了给Agent提供优质Context、把历史工作抽象为Workflow、去基模厂洗数据,剩下的高价值工作就是蹬Agent去探索“不在常识中”∩“难以迁移”的新威胁模型。
拍断大腿....
4月搓了个举一反三挖洞的agent harness,审了下kvm的arm64嵌套虚拟化的代码发现kvm_vcpu_init_nested对nested_mmus kvrealloc后不更新hw_mmu导致有UAF,后来忙各种内核洞应急就把这事儿彻底忘了
然后刚发现和6月v4bel的CVE-2026-46317撞了
1
29
2,796
"Sandbox as a tool wrapper" wins "Agent in sandbox"
Build long-running agents with more control over agent execution.
New capabilities in the Agents SDK:
• Run agents in controlled sandboxes
• Inspect and customize the open-source harness
• Control when memories are created and where they’re stored
962
C0ss4ck retweeted
Your Container Is Not a Sandbox
emirb.github.io/blog/microvm…
第一次被gvisor合pr
github.com/google/gvisor/pul…
最近在折腾高通的安全容器方案
1
3
982
最近工作吐槽记录
1. 对于一个修复流程完全不影响用户侧的固件漏洞,国内SRE讨论了一个多月“修复时监控哪些指标才能证明不影响用户”,非中SRE两周就修完了
2. 云产品没人相信APT就在面前,就算这些产品经理的前司都有公开的APT案例,也都装看不见...也没人相信APT会用bootkit,demo甩脸上也不相信...






















