Exploitation now follows disclosure within days, and next year's plan has to be built for it
Two critical flaws in the PHP core, dubbed php2shell, were fixed in emergency releases on 21 October. By 26 October, theUS Cybersecurity and Infrastructure Security Agency (CISA) had added both Known Exploited Vulnerabilities catalogue, its register of flaws confirmed as used in real attacks. Both are memory corruption bugs that could let an unauthenticated attacker run code on a web server, and both are rated 9.2 under version 4.0 of the CVSS scoring system. One security firm saw exploitation attempts against its honeypots on 21 October, the same day the fixes were released. Attackers didn't wait for anyone's change window.
The window is now measured in days, and not just for PHP
A VPN authentication bypass followed a similar pattern over a longer period. The vendor released a patch on 13 May, a security firm reported the first exploitation attempts on 17 May, and CISA listed the flaw on 29 May, sixteen days after the fix. Ransomware groups then used it to get in. Public reporting says the organisations that were hit had left the patch unapplied for months. Because the flaw bypassed authentication on the VPN itself, applying the patch was the control that decided the outcome.
Part of the speed comes from the tools attackers now use. On 13 November, the FBI warned that a range of threat actors have been using ARTEX, an open-source penetration testing framework driven by large language models, since at least August. It reports that some intrusions went from first targeting to privileged access in under three hours, and it singles out partner and supplier portals as weak points. A tool built for testers is now in attackers' hands, and it doesn't work to a quarterly cycle.
The underlying problem is a familiar one. Attackers move quickly because most of their process is automated. Defenders move slowly because theirs depends on people, and the slow part is rarely the patch itself. It is working out whether you run the affected component, who owns it, and whether anyone has the authority to take it offline on a Tuesday afternoon.
What the board actually needs to hear
We saw this play out with php2shell. When the flaws made the news, board members at one of our customers were rattled. They were convinced the company was about to be hit and wanted to know what was being done. The company doesn't run PHP. Nobody had told them, so the board was worried about a threat that never applied to the business.
That isn't a failing of the board. It is a communication gap. Directors hear about headline threats from the news, from peers and from the other boards they sit on, often before they hear from their own security team. If nobody inside the business gives them an answer, they will assume the worst.
Security leaders should get ahead of that. For any threat that makes the headlines, the board needs four answers, quickly and in plain language: are we affected, what is exposed, what are we doing about it, and when will it be closed. “We don't run PHP, so we are not exposed” is a complete and useful briefing. A CVSS score on its own is not.
How much technical detail should a board have? Less than many security teams fear, but more than none. Directors don't need to understand memory corruption. They do need enough fluency to ask the right questions and to recognise a vague answer, and with cyber risk now firmly on the board agenda, that is part of the job. A sensible split is that management owns the technical detail, while the board owns the questions, the deadlines and the exceptions.
What the plan should assume
A 2027 plan should assume that anything facing the internet will be attacked within days of a fix being announced. It should also assume you won't be certain where the affected component sits, and that some of what you run was inherited from an acquisition, a supplier or a team that has since moved on. How long a business stays exposed comes down to patching, end-of-life and legacy software, and privileged access. These weaknesses come up again and again in our audits and penetration tests, and the most common finding is old software that nobody has patched or retired. A plan that buys a new product and leaves them alone has spent the budget on the wrong problem.
That points to three commitments, each of which a board can measure.
An inventory of what you currently run. When a flaw turns up in a language runtime or a bundled library, the first question is where that component lives, so it tests your asset register before it tests your patching. It is also what lets you tell a worried board, within hours, that a headline flaw doesn't affect you. One large organisation has confirmed that an intruder accessed and took data from a legacy system it inherited through an acquisition. It hasn't said why the system was reachable. Inherited platforms often sit outside the main security programme after a deal, and that is exactly the gap an inventory should close. Every acquisition, every supplier-hosted application and every system older than the current team belongs in the register.
Patch deadlines in days for anything exposed to the internet. A monthly cycle is fine for internal workstations. It isn't fine for a VPN appliance, a web server, a source code platform or a remote management console. For those, Cyberlinx recommends a written standard: 72 hours for a known-exploited flaw and seven days for any other critical fix, with a board-approved exception process for cases that can't meet it. The CISA catalogue is a practical trigger. Once a product you run appears in it, the clock has started.
A named owner for every legacy system and third-party connection. The FBI's focus on supplier and partner portals is a governance issue as much as a technical one. If a connection has no owner, nobody is there to notice a supplier compromise, revoke access or ask whether the connection should exist at all. Every legacy system and supplier connection needs a named person who answers for it. Where nobody will take it on, record a decision to retire it, isolate it or accept the risk.
Who is most exposed?
Software and technology services businesses hold access to their customers' environments, so an attacker who reaches them can often reach their customers too. The FBI lists financial services and technology among the sectors it has seen targeted. Customers in those sectors often ask how quickly a flaw gets dealt with, so a slow patch becomes a commercial problem as well as a security one. Councils can face a similar exposure where public portals run on shared runtimes and depend on suppliers for patching.
What to do this week
1. Make a list of internet-facing assets. Run an external scan, compare the results with your asset register, and record every VPN, web server, source control platform and management console along with its version. Anything missing from the register is your first finding.
2. Set a patch deadline. Draft a patch standard for exposed systems (72 hours for known-exploited flaws, seven days for other critical fixes) and get it in front of whoever can approve it before the end of the year.
3. Put a name against every legacy system and supplier connection. If nobody will take one on, escalate it to the executive team as a decision rather than leaving it as an unowned task.
4. Agree who can pull a system offline. When there is no fix yet, someone needs the authority to take a system out of service without calling a committee together. Name that person, and the conditions, in advance.
5. Agree what the board hears, and when. Settle a short format for the next headline threat (are we affected, what is exposed, what are we doing, when will it be closed) and commit to sending it within 48 hours, including when the answer is that you are not affected.
Next step
If you are preparing a 2027 security plan for a board, book a planning session with us. We will work through the inventory, the patch standard and the ownership gaps with you, starting from what our audits and penetration tests keep finding.
Photo by Florian Steciuk on Unsplash
Related Articles








