blog

Security Operations

AI Agents

Security Tools

The Zero-Day Didn't Change. The Clock Did.

The Zero-Day Didn't Change. The Clock Did.

Yehonatan Rubin

|

|

Reading Time:

4 min

min

Table of Contents

On October 17, attackers started exploiting PHP2Shell. The patch shipped on October 21. That is four days of exploitation before anyone had a fix to apply.

PHP2Shell is a pre-auth remote code execution bug in PHP itself. Not in an app. Not behind a login. In the runtime under the app. A Censys snapshot from October 22 counted 36.7 million internet-facing services possibly exposed.

We have seen bad CVEs before. What changed is the clock. The window between disclosure and exploitation used to be weeks. Then days. Now exploitation can start before the disclosure. The old playbook assumed you had time to read the advisory, run a scan, open tickets and wait. That time is gone.

Four days before the patch

When exploitation starts before the patch, “patch faster” is not an answer. There is nothing to patch yet.

The only useful questions are about you. What do we actually run? Which of it is exposed? What can each exposed host reach? Who can act on it today?

Most teams cannot answer those questions in an afternoon. Not because they lack tools. Because the answers live in six tools and three people’s heads.

The legacy case is the normal case

PHP2Shell had a second twist. There is no standard fix for PHP 8.1 and older. Those versions are out of support.

That is not an edge case. That is most enterprises. Every large organization has systems that are well understood and too expensive to fix. They work. They make money. Nobody wants to rebuild them this quarter.

So the honest plan has two tracks. Cut the exposure now with compensating controls: a WAF rule, restricted access, a host pulled off the internet, rotated keys. Then track the upgrade as its own project, with its own owner and its own budget. Pretending the patch will come is not a plan.

The CVE is new. Your environment isn’t.

Here is the part I care about most as an engineer. Historical knowledge does not help much on day one of a new CVE. Your environment does.

At TReAT, the agents work a new CVE in four steps:

  • Connect. Match the CVE against what is actually deployed. The data comes from tools you already run: cloud security, scanners, code, ticketing, identity, WAF.

  • Prioritize. Rank each asset by exploitability, internet exposure, production use and what it can reach.

  • Plan. Patchable assets get the exact version and rollout order. Unpatchable ones get a compensating control.

  • Assign. Each action goes to its owner, in their own queue.

At one customer, the agents found 29 affected workloads from the cloud-security feed a day after the fix. Seven went to the top: internet-facing production. Most of those seven ran PHP 8.1. The plan blocked the exploit at the WAF and restricted access now, and opened the upgrade as a tracked project. Elsewhere, a PHP 5.4 host jumped the line. It was public, with cleartext cloud and SSH keys on it.

Nothing in that plan acted on its own. Each step went to a person to approve, and each one stayed reversible.

None of it needed a new workflow either. Same agents, same context. No project kickoff.

The real bottleneck was always ownership

Step four is where most zero-day responses quietly die. The finding is real. The plan is right. And the ticket lands with someone who left last year.

Acquisitions, regions and cloud sprawl leave findings with nobody to fix them. A finding with no owner is a knock on an empty house.

So TReAT rebuilds the ownership map every night from your own tools. Directory. Admin groups. On-call rotations. Repo teams. Who actually closed tickets in the last 90 days. An owner counts as confirmed only when two signals agree. Otherwise TReAT marks it likely, or asks you.

In the PHP2Shell case above, the owner took the work 34 minutes after TReAT opened it. That number is not about speed of typing. It is about the ticket reaching the right person the first time.

What a game plan looks like now

The next zero-day is already being written. You will not know its name. You can know everything else in advance. What you run. What is exposed. What each system can reach. Who owns it.

That is the shift. Zero-day response used to start at disclosure. Now it has to start the night before, with a map that is already right.

The CVE is new. Your game plan should not be.