惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

爱范儿
爱范儿
E
Exploit-DB.com RSS Feed
Google DeepMind News
Google DeepMind News
F
Full Disclosure
D
Darknet – Hacking Tools, Hacker News & Cyber Security
T
ThreatConnect
Stack Overflow Blog
Stack Overflow Blog
Last Week in AI
Last Week in AI
Martin Fowler
Martin Fowler
G
GRAHAM CLULEY
C
Check Point Blog
T
Threatpost
I
Intezer
Spread Privacy
Spread Privacy
The Register - Security
The Register - Security
Project Zero
Project Zero
月光博客
月光博客
人人都是产品经理
人人都是产品经理
阮一峰的网络日志
阮一峰的网络日志
D
DataBreaches.Net
IT之家
IT之家
Malwarebytes
Malwarebytes
T
The Blog of Author Tim Ferriss
P
Privacy International News Feed
P
Palo Alto Networks Blog
T
The Exploit Database - CXSecurity.com
量子位
李成银的技术随笔
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
Cisco Talos Blog
Cisco Talos Blog
Know Your Adversary
Know Your Adversary
美团技术团队
The GitHub Blog
The GitHub Blog
T
Tor Project blog
M
MIT News - Artificial intelligence
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Google Online Security Blog
Google Online Security Blog
P
Proofpoint News Feed
有赞技术团队
有赞技术团队
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
博客园 - 司徒正美
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
C
Comments on: Blog
T
Threat Research - Cisco Blogs
aimingoo的专栏
aimingoo的专栏
Security Latest
Security Latest
NISL@THU
NISL@THU
The Cloudflare Blog
H
Help Net Security
Recent Commits to openclaw:main
Recent Commits to openclaw:main

The Cloudflare Blog

The day my ping took countermeasures Announcing Claude Compliance API support with Cloudflare CASB Announcing Claude Managed Agents on Cloudflare Project Glasswing: what Mythos showed us Our billing pipeline was suddenly slow. The culprit was a hidden bottleneck in ClickHouse Browser Run: now running on Cloudflare Containers, it’s faster and more scalable When "idle" isn't idle: how a Linux kernel optimization became a QUIC bug Building For The Future How Cloudflare responded to the “Copy Fail” Linux vulnerability When DNSSEC goes wrong: how we responded to the .de TLD outage Code Orange: Fail Small is complete. The result is a stronger Cloudflare network Introducing Dynamic Workflows: durable execution that follows the tenant Post-quantum encryption for Cloudflare IPsec is generally available Agents can now create Cloudflare accounts, buy domains, and deploy Shutdowns, power outages, and conflict: a review of Q1 2026 Internet disruptions Making Rust Workers reliable: panic and abort recovery in wasm‑bindgen Moving past bots vs. humans Building the agentic cloud: everything we launched during Agents Week 2026 The AI engineering stack we built internally — on the platform we ship Orchestrating AI Code Review at scale Introducing the Agent Readiness score. Check to see if your site is agent-ready Shared Dictionaries: compression that keeps up with the agentic web Redirects for AI Training enforces canonical content Unweight: how we compressed an LLM 22% without sacrificing quality Agents that remember: introducing Agent Memory Agents Week: network performance update Introducing Flagship: feature flags built for the age of AI Cloudflare’s AI Platform: an inference layer designed for agents Building the foundation for running extra-large language models AI Search: the search primitive for your agents Deploy Postgres and MySQL databases with PlanetScale + Workers Artifacts: versioned storage that speaks Git Email for agents - Cloudflare Email Service now in public beta Project Think: building the next generation of AI agents on Cloudflare Introducing Agent Lee - a new interface to the Cloudflare stack Register domains wherever you build: Cloudflare Registrar API now in beta Browser Run: give your agents a browser Rearchitecting the Workflows control plane for the agentic era Add voice to your agent Managed OAuth for Access: make internal apps agent-ready in one click Securing non-human identities: automated revocation, OAuth, and scoped permissions Scaling MCP adoption: Our reference architecture for simpler, safer and cheaper enterprise deployments of MCP Secure private networking for everyone: users, nodes, agents, Workers — introducing Cloudflare Mesh Building a CLI for all of Cloudflare Durable Objects in Dynamic Workers: Give each AI-generated app its own database Agents have their own computers with Sandboxes GA Dynamic, identity-aware, and secure Sandbox auth Welcome to Agents Week 500 Tbps of capacity: 16 years of scaling our global network From bytecode to bytes- automated magic packet generation Cloudflare targets 2029 for full post-quantum security How we built Organizations to help enterprises manage Cloudflare at scale Why we're rethinking cache for the AI era Our ongoing commitment to privacy for the 1.1.1.1 public DNS resolver Introducing EmDash — the spiritual successor to WordPress that solves plugin security Introducing Programmable Flow Protection: custom DDoS mitigation logic for Magic Transit customers Cloudflare Client-Side Security: smarter detection, now open to everyone How we use Abstract Syntax Trees (ASTs) to turn Workflows code into visual diagrams A one-line Kubernetes fix that saved 600 hours a year Sandboxing AI agents, 100x faster Inside Gen 13- how we built our most powerful server yet Launching Cloudflare’s Gen 13 servers- trading cache for cores for 2x edge compute performance Powering the agents: Workers AI now runs large models, starting with Kimi K2.5 Introducing Custom Regions for precision data control Standing up for the open Internet- why we appealed Italy’s Piracy Shield fine From legacy architecture to Cloudflare One Announcing Cloudflare Account Abuse Protection: prevent fraudulent attacks from bots and humans Slashing agent token costs by 98% with RFC 9457-compliant error responses AI Security for Apps is now generally available Building a security overview dashboard for actionable insights Investigating multi-vector attacks in Log Explorer Translating risk insights into actionable protection: leveling up security posture with Cloudflare and Mastercard Fixing request smuggling vulnerabilities in Pingora OSS deployments Active defense: introducing a stateful vulnerability scanner for APIs Complexity is a choice. SASE migrations shouldn’t take years. From the endpoint to the prompt: a unified data security vision in Cloudflare One Ending the "silent drop": how Dynamic Path MTU Discovery makes the Cloudflare One Client more resilient A QUICker SASE client: re-building Proxy Mode How Automatic Return Routing solves IP overlap Always-on detections: eliminating the WAF “log versus block” trade-off Mind the gap: new tools for continuous enforcement from boot to login Stop reacting to breaches and start preventing them with User Risk Scoring Defeating the deepfake: stopping laptop farms and insider threats Moving from license plates to badges: the Gateway Authorization Proxy Evolving Cloudflare’s Threat Intelligence Platform: actionable, scalable, and ETL-less Introducing the 2026 Cloudflare Threat Report See risk, fix risk: introducing Remediation in Cloudflare CASB How Cloudy translates complex security into human action From reactive to proactive: closing the phishing gap with LLMs Modernizing with agile SASE: a Cloudflare One blog takeover Beyond the blank slate: how Cloudflare accelerates your Zero Trust journey The truly programmable SASE platform Toxic combinations: when small signals add up to a security incident We deserve a better streams API for JavaScript The most-seen UI on the Internet? Redesigning Turnstile and Challenge Pages ASPA: making Internet routing more secure Bringing more transparency to post-quantum usage, encrypted messaging, and routing security How we rebuilt Next.js with AI in one week Cloudflare One is the first SASE offering modern post-quantum encryption across the full platform Cloudflare outage on February 20, 2026
About the March 8 & 9, 2021 Verkada camera hack
Cloudflare Team · 2021-03-11 · via The Cloudflare Blog

Cloudflare uses a vendor called Verkada for cameras in our offices in San Francisco, Austin, New York, London and Singapore. These cameras are used at the entrances, exits and main thoroughfares of our offices and have been part of maintaining the security of offices that have been closed for almost a year.

Yesterday, we were notified of a breach of Verkada that allowed a hacker to access Verkada’s internal support tools to manage those cameras remotely, as well as access them through a remote root shell. As soon as we were notified of the breach, we proceeded to shut down the cameras in all our office locations to prevent further access.

To be clear: this hack affected the cameras and nothing else. No customer data was accessed, no production systems, no databases, no encryption keys, nothing. Some press reports indicate that we use a facial recognition feature available in Verkada. This is not true. We do not.

Our internal systems follow the same Zero Trust model that we provide to our customers, and as such our corporate office networks are not implicitly trusted by our other locations or data centers. From a security point of view connecting from one of our corporate locations is no different than connecting from a non-Cloudflare location.

While these camera devices were breached, it’s important to understand that inherently every employee has something like a root shell on a corporate network. It is simply not enough to rely on background checks, training on phishing and other threats or policies about not visiting questionable websites to keep malware off your network. This incident emphasizes the importance of the Zero Trust model that Cloudflare follows and provides to customers, which ensures that if any one system or vendor is compromised, it does not compromise the entire organization.

For example, a claim was made on Twitter that it might have been possible to access Cloudflare CEO Matthew Prince’s laptop.

This claim is spurious for a few reasons. Firstly, Matthew isn’t in the San Francisco office because of COVID-19. None of us are. So being on the corporate San Francisco network doesn’t get you closer to his or my machine.

Secondly, even if, somehow, an attacker were to get into his laptop they’d still have to get past the per-application, hard-key protected access control through our Zero Trust product Cloudflare Access.

And another claim was that the attackers had a “root shell inside Cloudflare’s network”.

That sounds scary but we don’t trust our corporate network; we use our products, such as Cloudflare Access, to control access to resources. The fact that the attacker had access to a machine inside the corporate network is no better than the kind of access they’d have had if they’d connected to our corporate WiFi network. The network isn't important, it's the access control that matters.

Of course, if we had been using the old castle-and-moat style of corporate networking (where anything and anyone on the corporate network are inherently trusted) the outcome could have been different.

This is why Zero Trust is so powerful. It allowed us all to work from home because of COVID-19 and it means that an attacker who got into the office network doesn’t get any further.

Internal Monitoring

Our monitoring and network intrusion detection system were able to capture the Verkada compromise and show exactly what network connections were made from the cameras. Additionally Verkada was able to provide us with a log of commands executed on our cameras.

Some press stories contain an image of the “root shell” on one of the cameras. It shows the attacker having executed commands like whoami, ifconfig and curl ifconfig.me.

Here are the commands executed by the intruder on March 8 (we are waiting for Verkada to supply us with the commands executed on March 9 but our other monitoring and logging tools give us a good sense of what those commands will be):

2358 Singapore       curl ifconfig.me

2357 New York        curl ifconfig.me

2352 San Francisco   curl ifconfig.me
2352 San Francisco   ifconfig
2352 San Francisco   whoami

1451 San Francisco   ifconfig
1451 San Francisco   whoami

Here’s a log of the traffic from the cameras. Let’s start with unusual DNS queries. This shows DNS lookups from two cameras in San Francisco, one in London and one in Singapore. Mostly the attacker was using ifconfig.me to find their external IP address.

In San Francisco they attempted to access Cloudflare’s main website, an admin panel that doesn’t exist (there is no admin.cloudflare.com) and a specific customer that I have replaced with example.com (the hacker also tried admin.example.com).

2021-03-09T00:01:15.373343Z	London	ifconfig.me
2021-03-09T00:01:15.373343Z	London	ifconfig.me
2021-03-09T00:01:15.372302Z	London	ifconfig.me
2021-03-09T00:01:15.371451Z	London	ifconfig.me

2021-03-09T00:01:00.580199Z	San Francisco	admin.example.com
2021-03-09T00:00:56.818201Z	San Francisco	admin.cloudflare.com.as13335.com
2021-03-09T00:00:56.768506Z	San Francisco	admin.cloudflare.com
2021-03-09T00:00:50.285755Z	San Francisco	ff.as13335.com
2021-03-09T00:00:45.014855Z	San Francisco	example.com
2021-03-09T00:00:45.011649Z	San Francisco	example.com
2021-03-09T00:00:33.418232Z	San Francisco	cloudflare.com
2021-03-08T23:59:49.677219Z	San Francisco	ifconfig.me
2021-03-08T23:59:49.677216Z	San Francisco	ifconfig.me

2021-03-08T23:59:27.261902Z	Singapore	cloudflare.com
2021-03-08T23:59:27.25919Z	Singapore	cloudflare.com
2021-03-08T23:58:30.274265Z	Singapore	ifconfig.me
2021-03-08T23:58:30.274155Z	Singapore	ifconfig.me

2021-03-08T23:52:23.145321Z	San Francisco	ifconfig.me

They made some of their requests using HTTP and we captured those requests

2021-03-09T00:01:15.375176Z London
ifconfig.me GET /
curl/7.71.1
200 OK

2021-03-09T00:01:00.630747Z San Francisco
admin.example.com GET /
curl/7.64.0
301 Moved Permanently

2021-03-09T00:00:45.041609Z San Francisco
example.com GET /
curl/7.64.0
301 Moved Permanently

2021-03-09T00:00:33.476684Z San Francisco
cloudflare.com GET /
curl/7.64.0
301 Moved Permanently

2021-03-08T23:59:49.708081Z San Francisco
ifconfig.me GET /
curl/7.64.0
200 OK

2021-03-08T23:58:30.289352Z Singapore
ifconfig.me GET /
curl/7.64.0
200 OK

2021-03-08T23:52:23.183201Z San Francisco
ifconfig.me GET /
curl/7.64.0
200 OK

We also have complete connection logging but in the interests of space I won’t put it all here.

The downloaded archived video

The attacker used the Verkada tools to download two archived recordings of our San Francisco office lobby. We only keep 30 days of video but when there’s a formal investigation we keep specific video for up to one year. The downloaded videos date back to July 13, 2020; we had kept them to support investigation into a theft. Here are some stills from each of those videos:

Timeline (UTC)

March 81451 - Unauthorized access to Verkada security cameras2352 - Unauthorized access is used to run curl ifconfig.me from a Cloudflare camera in San Francisco

March 90029 - Unauthorized download of an archived video from our front door camera in San Francisco0033 - Unauthorized download of an archived video from our elevator camera in San Francisco0040 - A second unauthorized download of an archived video from our elevator camera1750 - Contacted by a reporter about security incident involving cameras2107 - First internal report of a Twitter user claiming to have accessed our cameras with confirmation one of our cameras had been compromised2122 - Cloudflare disables all network connectivity of our security cameras2126 - Internal review of our corporate network security controls is completed2149 - Article is released about compromise of Verkada cameras