🚨 DirtyFrag (Linux kernel) — container → host root.
The public one-liner mitigation is incomplete. It blocks modprobe but leaves autoload-by-alias open.
Real fix needs all three layers:
cat >/etc/modprobe.d/dirtyfrag.conf <<'EOF'
install esp4 /bin/false
install esp6 /bin/false
install rxrpc /bin/false
blacklist esp4
blacklist esp6
blacklist rxrpc
alias net-pf-33 off
alias xfrm-type-2-50 off
alias xfrm-type-10-50 off
EOF
⚠️ Does NOT unload already-loaded modules. rmmod or reboot the node.
✅ Bonus: setting allowPrivilegeEscalation: false on the pod blocks the exploit at the container level. Verified.
securityContext:
allowPrivilegeEscalation: false
Safe for: HTTP, gRPC, DBs, kube-proxy, Istio, WireGuard, VXLAN. Breaks: strongSwan, Cilium IPsec, kernel AFS/RxRPC.
Pre-flight: ip xfrm state, ss -a -A rxrpc. Empty → ship it.
🔗 https://t.co/97mO7eMDBH
Post-training is where the magic happens for custom models, but the engineering overhead to train at the frontier is massive. With Baseten Loops SDK, we're solving the training-to-inference bottleneck. We've built an async RL and SFT framework that enables companies to train with RL at the frontier and pairs perfectly with our inference platform.
Sign up for early access: https://t.co/emom3pf1Qb
@orimanabu just curious, did they really verify the exploit script, after fix? alias net-pf-33 off
alias xfrm-type-2-50 off
alias xfrm-type-10-50 off
the alias needs to be off- verified
🚨 DirtyFrag (Linux kernel) — container → host root.
The public one-liner mitigation is incomplete. It blocks modprobe but leaves autoload-by-alias open.
Real fix needs all three layers:
cat >/etc/modprobe.d/dirtyfrag.conf <<'EOF'
install esp4 /bin/false
install esp6 /bin/false
install rxrpc /bin/false
blacklist esp4
blacklist esp6
blacklist rxrpc
alias net-pf-33 off
alias xfrm-type-2-50 off
alias xfrm-type-10-50 off
EOF
⚠️ Does NOT unload already-loaded modules. rmmod or reboot the node.
✅ Bonus: setting allowPrivilegeEscalation: false on the pod blocks the exploit at the container level. Verified.
securityContext:
allowPrivilegeEscalation: false
Safe for: HTTP, gRPC, DBs, kube-proxy, Istio, WireGuard, VXLAN. Breaks: strongSwan, Cilium IPsec, kernel AFS/RxRPC.
Pre-flight: ip xfrm state, ss -a -A rxrpc. Empty → ship it.
🔗 https://t.co/97mO7eMDBH
Inspect and filter every HTTP request leaving your microVM.
New post on @slicervm's proxy: secret injection without sentinels, OAuth that actually works, and stage-by-stage policy you can change mid-flight with code.
https://t.co/DfPPgLO9Pz
“中国制造”正从单纯的代工转向原创设计和高技术的“中国品牌”输出。徕芬(Laifen)就是一个例子。
这家总部位于中国深圳的科技公司,凭借高速吹风机在市场上成名,号称“戴森平替”。最近推出的P3 Pro 电动剃须刀,外观极具现代感、工艺精密、动力强劲,引起市场广泛关注。甚至被誉为“剃须刀界的劳斯莱斯”。
re: Vercel hack
This should be our wakeup call to get rid of API keys, just like we "got rid" of passwords with OAuth & passkeys
There's a fix. Of course there's a fix! It's called OIDC Federation, it works beautifully, but it still hasn't replaced API keys in the mainstream.
In short, your infra (Vercel, AWS, etc.) generates short-lived JWTs. You then tell your services (database, AI provider, etc.) to trust those JWTs instead of the API keys. No static secrets required.
These JWTs have a lifetime of <1h, and rotate automatically. Simpler + safer
Is it time?