Issue 13
Executive Signals | 20 November 2026
I have been watching the cybersecurity news with increasing amusement, and some concern. Every day, there seems to be another vulnerability discovered, another critical advisory issued, and another dramatic name attached to something that security teams are expected to investigate immediately. For some strange reason, we have developed a particular fondness for names ending in 2Shell. It has become something of a running joke.
I wrote about PHP2Shell in my previous Executive Signals edition. That was a serious vulnerability, and there was credible evidence of attackers exploiting it. Yet I find myself increasingly irritated by the seemingly endless variations of these names. I wonder how many security professionals feel the same way.
I have spent 34 years in the global technology industry, much of that time dealing with infrastructure, cybersecurity, operational risk and the consequences of technology failures. During those years, I have watched the industry respond to some extraordinary security incidents. We have gone from worrying about viruses spreading through floppy disks to malware capable of disrupting hospitals, shipping operations, industrial facilities and governments.
I remember a time when one significant computer virus could occupy the attention of the entire technology industry for weeks. A handful of major incidents in a year could dominate discussions among security professionals. Today, we have hundreds of vulnerability disclosures every day, and apparently every other one deserves its own name, logo, website and publicity campaign.
The obvious explanation is that the world has become much more dependent on software, and the number of vulnerabilities being discovered has grown enormously. Both are true. But I am worried about something else. We are creating an environment in which the sheer volume of warnings may eventually make it difficult for defenders to recognise the ones that really matter.
When a Computer Virus Was Big News
Let me go back a little. In November 1988, the Morris worm spread across the early internet, affecting an estimated 6,000 of the approximately 60,000 connected computers. The internet was still largely an academic and research network. Most businesses had little reason to worry about an internet worm because they were not connected to the internet in the first place.
The Morris worm was nevertheless a defining event. It demonstrated how quickly malicious or poorly controlled code could spread through interconnected systems. The incident helped accelerate the creation of the Computer Emergency Response Team at Carnegie Mellon University, which became an important part of the way the industry coordinated its response to security incidents.
By the time I began my career in the early 1990s, personal computers were becoming common in offices, networks were growing, and the threat from computer viruses was becoming a genuine operational concern.
In 1992, the Michelangelo virus became an international news story. It was designed to activate on 6 March, the birthday of the famous artist, and could destroy data on infected computers. There were warnings that enormous numbers of computers could be affected. People bought antivirus software, checked their systems and worried about what would happen when they turned on their PCs that morning.
The anticipated catastrophe never materialised on the predicted scale. There were real infections and some businesses suffered serious data loss, but the numbers were far smaller than many of the predictions. Contemporary reports debated whether the warnings had prevented widespread damage or whether the threat had been exaggerated.
That incident is worth remembering today. We were capable of creating a cybersecurity scare even before the internet became a household utility. Fear has always been an effective marketing instrument, and the security industry discovered its usefulness quite early. Michelangelo was still one major event, though, and for a while nearly everyone was talking about the same thing.
Towards the end of the 1990s came Melissa, followed by ILOVEYOU in 2000. Email had become an important business communication tool, and these attacks demonstrated how effectively human curiosity could be exploited. An attachment that appeared to come from somebody you knew was often enough to spread malicious code to other people.
Then came Code Red and Nimda in 2001, SQL Slammer and Blaster in 2003, Sasser in 2004 and Conficker in 2008. These were different kinds of attacks, using different methods of propagation and exploiting different weaknesses. Some spread with remarkable speed, while others remained persistent problems for years.
SQL Slammer was particularly impressive in a rather unpleasant way. In January 2003, it exploited a vulnerability in Microsoft SQL Server, infecting at least 75,000 hosts and reaching more than 90% of vulnerable hosts within ten minutes. A tiny piece of code was capable of causing extraordinary disruption. Those of us working in technology during that period saw how quickly an apparently technical problem could become a business problem. Networks slowed or stopped functioning, administrators scrambled to identify affected systems, and organisations discovered that their recovery procedures were not nearly as effective as they had assumed.
There were certainly many more vulnerabilities than the incidents I have mentioned. It would be inaccurate to suggest that the industry only had a dozen vulnerabilities to deal with every year. But the number of major events that commanded worldwide attention was relatively small, perhaps a dozen in a particularly difficult year. We knew which incidents had become major events because the consequences were already visible, or because the technical evidence suggested that an enormous number of systems could be affected. The publicity was not always proportionate, as Michelangelo demonstrated, but there was enough space between major events for organisations to investigate, respond and learn from them.
What surprised me, looking back through some of the contemporary research, was discovering that NIST had raised concerns about excessive security warnings as far back as June 2000. Even then, security professionals were worried about the frequency of announcements overwhelming organisations. Twenty-six years later, we are publishing more than 300 CVEs a day. We recognised the problem long before it reached its present scale.
From Computer Viruses to National Emergencies
As technology became more deeply embedded in business operations, the consequences of cyberattacks grew. Stuxnet, discovered in 2010, changed how many people thought about cybersecurity. It demonstrated that malicious software could interfere with industrial control processes and cause physical damage. For years, discussions about cybersecurity had largely concerned data, computer availability and financial fraud. Stuxnet showed what could happen when an attacker targeted the systems controlling machinery.
In 2014, Heartbleed exposed a serious weakness in OpenSSL, a component used extensively to secure internet communications. Later that year, Shellshock exposed vulnerabilities in the Bash shell. Both required significant attention because the affected software was widespread, often embedded in systems that organisations did not fully understand. The names were memorable, and the underlying weaknesses were serious enough to justify the attention they received.
Then came 2017. I suspect most cybersecurity professionals who were active at the time remember WannaCry. I certainly do.
On 12 May 2017, WannaCry began spreading internationally, exploiting a vulnerability in Windows systems. It eventually affected more than 230,000 computers in at least 150 countries. The United Kingdom’s National Health Service was particularly badly affected, with hospitals and medical facilities experiencing severe disruption. The UK National Audit Office later reported that at least 81 of England’s 236 NHS trusts had suffered disruption, either because they were infected or because they had to shut down systems as a precaution. Thousands of appointments and procedures were cancelled.
People needed medical treatment, hospitals needed access to patient records, and healthcare professionals needed systems that had suddenly become unavailable. Nobody was arguing any more about whether a vulnerability should be classified as critical or high. The consequences were visible to patients, doctors, hospital administrators and government officials. The world reacted. Security teams examined their systems, vendors issued guidance, organisations accelerated patching and governments became involved in responding to the incident.
Just six weeks later, NotPetya demonstrated another dimension of the problem. The attack entered through software used for filing tax in Ukraine and spread into the operations of major international businesses. Maersk, one of the world’s largest shipping companies, suffered widespread disruption. Applications and data became unavailable, systems were shut down, and employees resorted to manual procedures to keep operations moving. Maersk estimated that the attack would cost it between $200 million and $300 million. For a company moving goods across the world, the inability to operate its computer systems was also an inability to perform much of its business.
WannaCry and NotPetya were two separate incidents, but they forced management teams to reconsider some uncomfortable assumptions about the resilience of their organisations. Cybersecurity could no longer be treated merely as a technical problem for the IT department. It had become a matter of business continuity, operational survival and, in some cases, public safety.
In December 2021, Log4Shell brought the industry another major emergency. A vulnerability in a commonly used Java logging library meant that an enormous number of products and applications had to be examined. Many organisations did not even know which of their applications contained the affected component. Beyond the severity of the vulnerability, the extent of exposure was what nobody could see. Security teams had to identify vulnerable software buried inside applications, appliances and third-party products, often without a reliable inventory of those dependencies. This is the kind of vulnerability that justifies an extraordinary response. Its reach was enormous, exploitation was practical, and security teams needed to act quickly.
| Year | Event | Scale |
|---|---|---|
| 1988 | Morris worm | About 6,000 of roughly 60,000 connected computers |
| 1992 | Michelangelo | Predicted far more damage than occurred |
| 2003 | SQL Slammer | At least 75,000 hosts, over 90% of vulnerable hosts within ten minutes |
| 2010 | Stuxnet | Malicious software interfering with industrial control processes |
| 2014 | Heartbleed and Shellshock | Weaknesses in OpenSSL and the Bash shell, both widely embedded |
| 2017 | WannaCry | 230,000+ computers in at least 150 countries. At least 81 of 236 NHS trusts in England disrupted |
| 2017 | NotPetya | Maersk estimated $200 to $300 million in losses |
| 2021 | Log4Shell | Vulnerable component buried in countless applications, often unknown to their owners |
Now compare the attention these incidents received with the amount of vulnerability information being published today.
The Industry Has Become a Vulnerability Factory
According to research published by Black Kite on 18 November 2026, the number of CVEs published this year had already reached 100,009. The whole of 2025 saw 48,150. That works out to approximately 311 new CVE records every day in 2026, one every 4.6 minutes. October alone produced 17,051.
Published CVEs · Black Kite, 18 November 2026
Bars drawn to scale. Published CVE records, not vulnerabilities exploitable in any one organisation.
I find these numbers astonishing. A dozen years ago, we were discussing individual viruses and worms that made international headlines. Today, an organisation can receive more vulnerability information in a morning than its security team can meaningfully investigate in a week.
Of course, not every published CVE represents a new, actively exploitable weakness. Some concern obscure products or configurations, some document older problems, and some are the result of improvements in research and reporting. The rapid growth in 2026 has also been influenced by automated and AI-assisted discovery, an increase in organisations authorised to assign CVEs, and the publication of previously undocumented open-source vulnerabilities. Those are useful developments in their own right. I would rather have researchers discover and disclose weaknesses than leave them available for attackers to exploit without our knowledge.
The problem begins when we treat the publication of a vulnerability as though it automatically constitutes an operational emergency for every organisation. A vulnerability might receive a CVSS score of 9.8 or 10.0, but that tells me very little about the actual exposure of a particular business.
What a CVSS score does not tell you
Does the organisation use the affected software?
Is the vulnerable component installed and enabled?
Can an attacker reach it?
Are there controls preventing exploitation?
What would happen if the component were compromised?
These questions require understanding the organisation and its technology environment. A severity score or a headline announcing the vulnerability cannot answer them. Yet the volume of notifications keeps growing. Every vendor has an advisory service. Every threat intelligence provider wants to alert its customers. Cybersecurity publications compete to report new discoveries. Social media adds another layer of commentary, speculation and sometimes outright nonsense.
The naming convention has become particularly entertaining. Everything seems to be turning into another variation of Something2Shell. We have reached a point where the name sometimes receives more attention than the technical conditions required to exploit the vulnerability.
I am not suggesting that researchers should stop giving vulnerabilities memorable names. Log4Shell was a useful shorthand for a complicated and serious problem. Heartbleed became widely recognised among people who would otherwise never have heard of OpenSSL. But when every other announcement begins to look like a major historical vulnerability, we start losing the ability to distinguish one from another.
I would not be surprised if the next major vulnerability came with a promotional video, a merchandise collection and somebody offering a commemorative coffee mug. Perhaps that is an exaggeration. Looking at some of the marketing material produced by the cybersecurity industry, I am not entirely sure.
The People Who Have to Defend All This
I have worked with and led technology and security teams long enough to know that vulnerability remediation is rarely as simple as downloading a patch and installing it. In an enterprise environment, somebody has to determine whether the vulnerability affects the organisation. Somebody must identify the affected systems, verify their configurations and understand the dependencies. Patches may require testing, application owners may need to approve changes, and business teams may refuse downtime because production cannot be interrupted.
There may be legitimate reasons to delay a patch. There may also be poor reasons, including inadequate inventories, unclear ownership, obsolete applications and an unwillingness to spend money on maintenance. All of them exist, sometimes within the same organisation.
A security team can identify a critical vulnerability on Monday, recommend remediation on Tuesday and still be discussing the change window two weeks later. During those two weeks, another few thousand CVEs may have been published. Initially, the team investigates every warning. Then it begins prioritising. Eventually, it develops a routine for handling the constant flow. Tickets are created, reports are generated, exceptions are documented, and critical vulnerabilities become another category in the backlog.
I have seen how easily management can become comfortable with these numbers. A report might show 1,000 critical vulnerabilities, of which 700 have been remediated. The organisation has achieved a 70% closure rate, and everyone is pleased with the improvement.
A report that looks good
A 70% closure rate. Which of the remaining 300 gives an attacker a direct route into the most important business systems?
Illustrative figures from the example above, not survey data.
What if one of the remaining 300 provides an attacker with a direct route into the organisation’s most important business systems? What if another allows an unauthenticated attacker to compromise an internet-facing device? And what if the team has no reliable way to distinguish those weaknesses from hundreds of others that are not practically exploitable in its environment? I would be much more interested in those questions than the percentage of tickets closed.
We measure security activity more easily than we measure the reduction of actual exposure.
The constant stream of vulnerability announcements is adding to an operational problem we have already struggled with for years. It has consequences for the people doing the work. Security professionals have finite attention and limited time. The infrastructure and application teams responsible for remediation are also doing other work. They have systems to maintain, projects to deliver, incidents to investigate and business operations to support. We can automate vulnerability discovery and generate findings continuously, but we cannot assume that the organisation’s ability to investigate and remediate those findings will increase at the same rate.
There is a human cost to repeatedly telling people that everything is urgent. After a while, urgency becomes part of the background noise. I suspect we are already seeing some of that behaviour, although I would be careful about attributing poor remediation performance entirely to fatigue. Weak governance, technical debt, inadequate staffing and business priorities all contribute. What concerns me is that the growing volume of warnings can make those existing problems considerably worse.
AI Is Going to Add More Fuel
We have now reached the point where AI systems can assist researchers in identifying vulnerabilities, analysing source code and investigating potential exploitation paths. Some of the results are impressive. Weaknesses that survived years of conventional analysis are being identified through new approaches. I welcome that development. Better research should eventually produce better software.
But I am not convinced that we have fully considered what happens when the discovery process becomes much faster while the remediation process remains constrained by the realities of running businesses. Suppose AI-assisted tools become capable of discovering thousands of additional vulnerabilities every week. The security industry will celebrate the increase in research productivity, and reasonably so. But somebody still has to determine which findings matter, assess whether affected software is deployed, decide what should be fixed and coordinate the necessary changes. We could find ourselves in the peculiar position of having vastly better vulnerability intelligence and still being unable to improve the security of the systems we actually operate.
Attackers have a different problem. They do not need to process every vulnerability affecting every system. They need to identify weaknesses that are useful against their chosen targets. They can focus on publicly exposed services, commonly deployed products, weak identities and the routes that allow them to move through an environment. AI-assisted reconnaissance, vulnerability analysis and exploitation may help attackers conduct that work faster. There is already evidence of AI being used in parts of offensive operations, although I would not accept every claim of fully autonomous cyberattacks without examining the supporting evidence.
The imbalance
An enormous inventory of vulnerabilities across thousands of systems. Every finding has to be assessed, scheduled, tested and fixed, in a business that is still running.
A much smaller number of exploitable paths: exposed services, common products, weak identities, and the routes through an environment.
That imbalance is what interests me. We should be investing at least as much effort in understanding real exposure and validating remediation as we are investing in discovering more vulnerabilities.
When the Wolf Really Comes
There is a reason the story of the boy who cried wolf has survived for centuries. Lying is only part of it. The villagers initially responded because they believed the warning and understood the danger. Repeated false alarms eventually changed their behaviour. When the wolf finally arrived, the warning no longer produced the necessary response.
Our cybersecurity problem is slightly different. Most vulnerability warnings are not false. The underlying flaws may be genuine, and the researchers reporting them may have done excellent work. But every vulnerability is not an equal threat to every organisation, and every newly announced weakness does not deserve the same response.
We have experienced genuine industry-wide emergencies before. Morris, Melissa, ILOVEYOU, SQL Slammer, Stuxnet, Heartbleed, WannaCry, NotPetya and Log4Shell each gave us reasons to reconsider how we protected our systems. The response to those events was not always perfect, but the industry generally recognised that something significant had happened.
Imagine the next major vulnerability being announced in the middle of a day when another 300 CVEs have already appeared. It affects software used by thousands of organisations, can be exploited remotely, and gives attackers access to systems supporting critical business operations. The announcement gets a dramatic name. Vendors publish emergency advisories. Researchers demonstrate exploitation. Threat intelligence feeds light up, and social media fills with people explaining why this is the most dangerous vulnerability of the year.
Meanwhile, a security team has several hundred other critical findings in its backlog. The infrastructure team has maintenance scheduled for the following weekend. The application owner wants more testing. Management believes the vulnerability is being handled because a ticket has been raised and assigned. None of these people is necessarily negligent or incompetent. They may simply be responding to the latest warning in exactly the same way they have learned to respond to hundreds of earlier warnings. That is the situation I worry about.
Having spent 34 years watching technology evolve, I am not pessimistic about our ability to build more secure systems. We have considerably better tools, stronger security practices and far more capable professionals than we had in the early 1990s. The internet itself is vastly more resilient than it was during the early days of widespread worms and viruses. But the complexity of the systems we operate has grown enormously, and the number of warnings competing for the attention of their defenders has grown with it.
I would like to see security leaders, vendors and researchers give more attention to communicating the conditions under which a vulnerability becomes genuinely dangerous. We need to know what is exposed, whether it is exploitable, what an attacker can achieve and which systems require immediate attention. A CVSS score and a frightening headline are useful starting points, but they cannot substitute for that assessment.
Executives should also recognise that a growing vulnerability backlog is not merely a problem of security teams failing to patch quickly enough. It may reflect an organisation that has never properly understood its technology estate, its business dependencies or the consequences of allowing an attacker to move between them.
I have no doubt that another major cyber vulnerability will eventually appear. It could affect infrastructure used by almost every organisation, or a widely deployed technology that few people realise they depend upon. It could be exploited much faster than defenders can reasonably patch it. When that happens, the industry will sound the alarm. There will be no shortage of people shouting about it.
My concern is whether the people responsible for defending our systems will still be able to recognise the difference between that warning and the hundreds they received the day before.
The closing signal
After all, the wolf doesn’t have to fool everybody. It only has to find one village that has stopped paying attention.
Sources and Further Reading
- FBI, The Morris Worm. The 1988 worm and its reach.
- ICS-CERT, Stuxnet Indicators, September 2010.
- Institute for Science and International Security, Stuxnet Malware and Natanz, December 2010. Analysis of its effects on Iranian centrifuges.
- Black Kite, 100K CVE Report, 18 November 2026. These are published CVE records, not a count of vulnerabilities exploitable in any one organisation. The daily average and publication interval are calculated from the reported total.
- CISA, Mitigating Log4Shell and Other Log4j Vulnerabilities, December 2021.
- CISA joint advisory AA21-356A, archived PDF.
- CAIDA, The Spread of the Sapphire/Slammer Worm, 2003.
- NHS England, Lessons Learned Review of the WannaCry Ransomware Cyber Attack, 2018. Cites Europol’s estimate of more than 230,000 computers in at least 150 countries.
- UK National Audit Office, Investigation: WannaCry cyber attack and the NHS, 27 October 2017.
- NIST, Mitigating Emerging Hacker Threats, ITL Bulletin, June 2000. Melissa, ILOVEYOU, and concern about the volume of security warnings.
- CERT/CC, Heartbleed, Vulnerability Note VU#720951.
- Red Hat, Shellshock vulnerability.
- Carnegie Mellon University Software Engineering Institute, History of cyber incident management. DARPA’s request to establish CERT/CC after the Morris worm.
- The Washington Post, Michelangelo PC virus wasn’t quite an epidemic, 7 March 1992.
- A.P. Moller-Maersk, Q2 2017 Interim Report, 16 August 2017. NotPetya and the estimated financial impact.
