The leader in real-time cloud security
San Francisco, CA
Joined January 2014
- Tweets7.8K
- Following1.3K
- Followers10.4K
- Likes4.3K
The board question on AI is coming. Get your answer before November 17. ⏳
Frank Dickson of Dickson Research and Sysdig’s Blair Howard cover the five questions every security leader should be able to answer, in 30 minutes.
📅 Tue, Nov 17 | 8am PT / 11am ET
okt.to/cNCtwV
#AIDefense #AIAgents #CISO
1
1
178
Every pickle defense in machine learning has been bypassed the same way. 🥒
Sysdig TRT’s Lydia Graslie analyzed six CVE bypasses across PyTorch, Keras, and picklescan and found the same flaw: the checker and the loader disagreed about what the file was. A pickle isn’t data. It’s a program.
okt.to/4vj7AJ
#CloudSecurity #ThreatResearch #MLSecurity #RuntimeSecurity
When an agent read a credential file, its own log had no entry 82% of the time. The kernel logged every read. 🔒
One organization, one month of AI agent activity, and what Sysdig AI Defense found.
okt.to/GAuFp5
#AIDefense #AIAgents #RuntimeSecurity
We talk a lot about speed. The Milan team decided to run it. 🏃
Two teams of four took on The Loop Relay, a 20K relay with skyscrapers, team spirit, and zero latency on the cheering.
Shoutout for lacing up.
Want to join this team? okt.to/HYlnIT
#LifeAtSysdig #TeamSysdig #Milan
Detect. Respond. Accessorize. 🧢
Pick up your Lumin cap at booth 168 at KubeCon SLC, and drop into our Falco sessions:
📅 Nov 11, 10:30am | Booth 168: Falco Meets AI Threats (live demo)
📅 Nov 12, 2:40pm | Salt Palace 251 D-F: Falco Never Flies Alone
KubeNight Out: okt.to/TSP1K3
#KubeConSLC #Falco
An agent can tell you it did one thing. The real story might be something else. 🔒
Today, Sysdig is introducing AI Defense: runtime security and governance for agents. It finds every agent your team runs, decides what they’re allowed to do, and keeps a record the agent can’t change.
okt.to/s9AfEg
#AIDefense #AIAgents #RuntimeSecurity
Last chance to register: Thursday’s LinkedIn Live is almost here. ⏳
Can you trust an AI agent’s own account of what it did? Sysdig’s Conor Sherman, Shantanu Gattani, and Ryan Davis take it on live, with a demo.
📅 Thursday, Nov 5 | 12pm ET / 9am PT
okt.to/il8UKv
#AIDefense #AIAgents #RuntimeSecurity
About 50% of AI tools on employee laptops never show up in the agents’ logs. 🔒
Frank Dickson of Dickson Research and Sysdig’s Blair Howard take on the five AI questions every security leader should be able to answer.
📅 Nov 17 | 8am PT / 11am ET
okt.to/ewzMrm
#AIDefense #AIAgents #CISO
From Raleigh to Milan, the Sysdig team went all out for Halloween. 🎃
Masquerading is usually something we detect. This week, we celebrated it. Zero false positives. Several very high-severity outfits. 👻
Want to join this team? okt.to/zHFr7D
#LifeAtSysdig #TeamSysdig #Halloween okt.to/cP4dCZ
Sysdig Headless Cloud Security won the Agentic AI Security Innovation Award in the 2026 CyberSecurity Breakthrough Awards. 🏆
In an industry competing on dashboards, we broke through by removing them.
okt.to/BfXzOs
#CybersecurityBreakthrough #HeadlessCloudSecurity #AIAgents #Sysdig
🚨 Three bugs. One week. The new patch window is shorter than the news cycle.
🐘 PHP2Shell: unauthenticated RCE in PHP core, exploitation confirmed
🟢 Node2Shell: Node.js Brotli bug, also being exploited
🐧 ZDI-CAN-33367 and five other Linux flaws, one with a public root exploit
Vulnerability scanning and runtime detection are best friends before patches land. Blog links in the replies. 👇
#CloudSecurity #ThreatResearch #PHP #NodeJS #Linux
PHP2Shell
sysdig.com/blog/php2shell-cv…
Node2Shell
sysdig.com/blog/node2shell-c…
ZDI-CAN-33367
sysdig.com/blog/detecting-an…
1
210
Mythos access is expanding. Your vulnerability management program doesn’t need a rewrite. 🔒
More defenders can find vulnerabilities faster, but discovery isn’t exploitation. The fundamentals still hold.
Blog: okt.to/oSDBrm
Last chance to register for “Mythos, Decoded” this Thursday, Oct 29 at 8am PT / 11am ET: okt.to/5UTR1P
#CloudSecurity #VulnerabilityManagement #AIAgents okt.to/ZRPbfw
🚨 Node2Shell (CVE-2026-111497, CVSS 9.2): a public repo for this Node.js Brotli flaw appeared about five hours after the fix shipped.
No confirmed in-the-wild exploitation yet, but the patch-to-PoC gap was about a day. Patch in order of reachability: start with services that decode Brotli you don’t control.
okt.to/3pTBzk
#CloudSecurity #ThreatResearch #NodeJS #RuntimeSecurity
🚨 A working root exploit for a Linux kernel bug is public. No CVE. No merged fix.
ZDI-CAN-33367 is a gap in the Dirty Frag fix, with root confirmed on current Ubuntu, Debian, and AlmaLinux. Restrict unprivileged user namespaces and enforce seccomp RuntimeDefault while you wait.
okt.to/08psRf
#CloudSecurity #ThreatResearch #Linux #RuntimeSecurity
AI agents are writing code, touching production, and acting with real credentials. So who’s watching what they actually do? 🔒
Join Sysdig’s Conor Sherman and Shantanu Gattani on LinkedIn Live to talk runtime security and governance for AI agents. Live demo included.
📅 Nov 5 | 12pm ET / 9am PT
okt.to/1h0YSs
#AIDefense #AIAgents #RuntimeSecurity
🚨 PHP2Shell: two critical CVEs in PHP core (CVSS v4.0 9.2). Exploitation attempts began four days before the patch was released.
An unauthenticated attacker can run code with HTTP requests alone. The vulnerable code dates to PHP 5.4 in 2012, so even clean application code is exposed.
Patch first and assume breach.
okt.to/xlQvRe
#CloudSecurity #ThreatResearch #PHP #RuntimeSecurity
Wide eyes. Racing pulse. A backlog that keeps growing. You may be experiencing Mythos anxiety. 😅
Join our “Mythos, Decoded” webinar to learn how runtime context shows which vulns add real risk. Practical steps, minus the panic.
📅 Oct 29 | 8am PT / 11am ET
okt.to/vOdlGe
#CloudSecurity #VulnerabilityManagement #RuntimeSecurity
1
192
Sysdig Accelerate ‘26 Japan is a wrap. 🙌
Thank you to our speakers, customers, and partners. The conversations in the room were exactly why we do this: people serious about defending cloud and AI environments at machine speed.
See you next year, Japan. 💚
#Sysdig #CloudSecurity #AIDefense #Japan okt.to/LvyX7Q
Frontier AI models like Mythos can discover vulnerabilities faster than any human. Most advice floating around isn’t helping.
Join us October 29 to separate what’s actually changed from the noise.
Mythos, Decoded: What’s Actually Changed in Vulnerability Management
📅 Oct 29 | 8am PT / 11am ET
okt.to/iFTNAp
#CloudSecurity #VulnerabilityManagement #AIAgents