On Naming the Flaws in Our Machines
Why are we naming security vulnerabilities and what is the impact of doing so?
This post is more an experiment than anything else. I've been wanting to expand my ability to write for a while now, adding some sort of 'literary touch' to my usual sober, analytical approach. Please keep this in mind when reading it.
In the second chapter of Genesis, before anything else is asked of him, Adam is brought the animals "to see what he would call them". Naming comes before tending, before toil, before the fall. It is the first thing a person does to the world, and it's not innocent. To name something is to pull it out of the undifferentiated background and to say that this is a thing, something distinct from the others, something to be spoken of.
For most of its history the software industry declined to do this with its own failures. A flaw in a program was a number, with identifiers such as MS08-67, or a string such as CVE-2024-0160.
If you recognize and / or remember MS08-67, congratulations on the receding hairline. And don't forget the Ibuprofen for your back and scheduling your prostate exam.
These are not names in any 'human' sense. They are more akin to catalogue numbers of a museum archive or a library, useful for retrieval and (rather) useless for memory. Yes, malware did occasionally have names, but malware is something that does things. Vulnerabilities, on the other hand, are like conditions. And conditions were left anonymous.
That changed in April of 2014, when a small security firm from Finland found a flaw in OpenSSL. Because the bug in question let a server 'bleed' out its memory through a feature called "heartbeat", an employee of the company named it "Heartbleed". Following this, a designer quickly drew a red heart dripping blood and the company bought a domain, using it to publish a plain-language explanation of the bug.
When I wrote the initial draft of this post it ended up being a story about marketing, about hype, about communication and the negative impact all of this has had on the cybersecurity industry.
Yet the more I wrote and the more I thought about it, the more I realized that instead of publishing just another rant about some of the more annoying aspects of naming vulnerabilities, I want to tell a story about naming itself - what happens to a thing when it's given a name? What does a name invoke in or ask of those who hear it? And what about the things that remained unnamed?
Infrastructure is, by its nature, (mostly) invisible, only becoming visible upon breakdown. Only when it breaks does it step forward as an, for lack of a better word, object demanding attention. Cryptographic libraries are infrastructure in the purest sense.
Almost no one who used the Internet in early 2014 knew that OpenSSL existed, that it was maintained by barely a handful of volunteers, or that it was receiving around two thousand dollars a year in donations while protecting the digital commerce of the world; "Heartbleed" was the breakdown.
However, with digital infrastructure like a programming library (no matter how popular or widespread it might be), the breakdown is noticed, at first, by almost nobody. Its failure is silent, abstract, distributed across millions of machines that keep on working despite it.
This is the first and perhaps deepest thing a name did. A CVE number (or any other technical identifier) makes a breakdown recordable; a name makes it perceptible. When Heartbleed happened, a survey found that nearly two-thirds of American Internet users had heard of it, and many of them changed their passwords because of it.
Most of them could not have explained what a buffer over-read was. They did not need to. They had been shown, for the first time, that there was a hidden structure underneath their digital daily lives, and that it could bleed.
I have written and re-written the last sentence at least a dozen times, trying to make it sound less cheesy and over-the-top, but no matter what I tried, I still ended up with this. I'm genuinely sorry for that cheese.
What followed was not only the obligatory mitigating and patching, but a kind of moral reckoning with the invisible. Within weeks, the Linux Foundation had assembled a fund for the neglected projects on which everyone depended. Seven years later, a flaw in the logging library Log4j (named, aptly, "Log4Shell"), led to a White House summit on open-source security and the first inquiry of a new US Cyber Safety Review Board.
In both cases the scandal was not only the technical security issue. It was the discovery of how much had been, and continues to be, resting on how little, and on how few people.
The name was the occasion for that discovery. It turned an infrastructure failure into a public fact, and public facts can be argued over, funded and governed in a way that archive entries cannot. And, going back to the fact that naming does something to thing it describes, a named vulnerability is treated differently from an unnamed one. Its significance is not fixed by its technical properties alone. It is partly constituted by the response it provokes.
Executives ask about it by name. Customers send questionnaires asking whether a company is affected by it, or how it's dealing with it. Vendors publish dedicated advisory pages. Journalists follow up. Attackers, too, read the news, and famous bugs sometimes draw exploitation partly because they are famous.
It's not just a name results in a different treatment, the name itself produces a difference, like the urgency that seemed to justify the name in the first place. There is no neutral fact of the matter about how much attention a vulnerability deserves, waiting to be read o before the naming happens.
The act of naming enters into the reality it describes. Heartbleed was serious, but it also became, through its name, the kind of event that reorganizes funding and policy. An equally serious flaw that remained a number might never have become that kind of event at all.
A major difference between naming 'traditional' objects and vulnerabilities is that traditionally, names do rarely tell you anything about it; taking "Dartmouth" as an example, the town would remain Dartmouth even if the river Dart changed its course. Vulnerability names are the opposite, they are engineered to connote. They are not (just) labels attached to a thing, but interpretations of it.
If we accept that naming confers power over how a thing is perceived, then the question of who names becomes a question of authority, and of credit.
A lot of people tend to have forgotten that the vulnerability that eventually became Heartbleed was found first by a Google engineer named Neel Mehta, two weeks before Codenomicon named it. People remember the company and the logo, not the Google engineer.
In the older 'economy of discovery', fame and glory went to the discovered. One spoke of the "Kaminsky flaw" after the researcher who found a severe weakness in the domain name system in 2008. In the new economy, glory went to whoever named, and the name could be owned in a way the discovery could not.
This isn't a new pattern. Explorers named continents after themselves and their patrons, regardless of who had lived there first or arrived sooner. To name is to annex. The proliferation of vulnerability brands after 2014, with their domains, logos and press kits, was in one sense a land rush over a newly discovered territory: public attention to security.
The (so far) darkest version of this rush came in 2018, when a little-known firm published a set of branded flaws in AMD processors on a dedicated website after giving the company less than a day's notice, while a short-selling research outfit simultaneously published a report predicting AMD's collapse.
The flaws were real but required administrator access to exploit. What was on display was the name used not to illuminate a danger but to move a price. Here the act of naming had become fully detached from its protective justification. It was rhetoric in the oldest and most suspect sense, speech aimed at an effect rather than truth.
Yet the authority to name has also been democratized in ways that cut the other direction. In 2025, the researcher Kevin Beaumont dubbed a new Citrix flaw "CitrixBleed 2", explicitly linking it to a predecessor, while the vendor was still saying there was no evidence it was being exploited. The US cybersecurity agency soon listed it as actively exploited.
In this case the name was a tool of the defenders, not of the discoverers or the vendor. Naming was used to contest an official reassurance, the way a community gives a name to a scandal that its authorities would prefer to leave unspoken.
While I'm really not a big fan of quoting Chinese philosophers in the context of cybersecurity, I have to use Confucius to make a point.
Just for the record: I'm going to yell at the next person who tries to apply "The Art of War" to dealing with APTs. Infosec community, just stop.
Confucius, asked what he would do first if he were given the government of a state, answered that he would rectify the names. If names are not correct, he said, speech does not accord with reality; if speech does not accord with reality, aairs cannot be carried to success.
For Confucius, the corruption of names was the root of disorder: when a ruler is not truly a ruler but is called one, or a minor thing is called great, the whole structure of action built on those words begins to fail.
The security industry rediscovered this problem the hard way. Once a logo became the sign of importance, every researcher had an incentive to acquire one, and the sign began to drift away from the thing it signified. It felt like we, as an industry or a community, have entered some sort of branding arms race, a race in which each finding had to sound bigger and scarier than the last.
The emblematic case was something called "Badlock" in 2016, announced weeks in advance with a logo, a website and a countdown, which turned out to be a moderate flaw that Microsoft rated "Important" rather than critical. The community's annual Pwnie Awards gave it the prize for most over-hyped bug.
Badlock was a failure of the rectification of names. It was called a great thing and was a modest one. The cost of such failures is not only embarrassment; it is the slow debasement of the currency. If every named bug is a catastrophe, then "named bug" stops meaning catastrophe, and the next real one must shout louder to be heard. If you're reminded of parallels about some shepherd boy and a wolf, you're right.
And the community responded to this, in their own, charming way. There was no regulator of names, no standards body, no licensing authority. Instead there was ridicule.
Parody websites appeared. In 2018 a researcher published a deliberately absurd branded page for a privilege escalation in beep , a utility that makes a computer emit a beep, complete with its own name, "Holey Beep", and it too won the over-hype prize. Satire functioned as a distributed rectification of names: a way for a community to remind itself that the sign must answer to the thing.
One institution did, however, try a more systematic answer to the problem. In 2020, CERT/CC, one of the oldest coordinators of vulnerability disclosure, launched a social media bot that would automatically give every vulnerability a randomly generated, deliberately meaningless name, an adjective and a noun, in order to remove fear from the process. If names were going to exist, perhaps they could at least be stripped of rhetoric?
The experiment did not take hold, and its failure is instructive. A random adjective and a random noun still mean something. Language does not permit true neutrality. To refuse a meaningful name is not to escape meaning, it is only to let chance choose it.
So the choice was never between naming and not naming, or between loaded names and neutral ones. It was between names chosen carefully and names chosen carelessly, and between names chosen by people with an interest in the truth and names chosen by people with an interest in something else.
Every act of naming leaves something in the shadows, and this applies to vulnerabilities as well. Some flaws are framed as worth caring about, and others are not, with the frame often being set before anyone measures the damage.
People who have to deal with computer intrusions are well aware of this. Heartbleed quickly vanished from security assessments or incident reports, while "boring " Windows flaws like MS08-067 and MS17-010 kept turning up everywhere. These are numbers, not names.
One of them was the weakness that let the "WannaCry" ransomware spread around the world in 2017; the exploit and the worm both got names ("EternalBlue", "WannaCry"), while the hole they passed through stayed a number. The unnamed vulnerabilities are the dark matter of the industry: the overwhelming majority, invisible, persistent, unglamorous.
Here the name's great virtue, its power to make the invisible visible, becomes the source of its great vice. Attention is finite. I'm sure there's a professional way to word this, but I'd argue that a wealth of information in turn often creates a poverty of attention.
Every name that gathers attention draws it away from somewhere else, and there is no reason to believe that the somewhere else is less dangerous. The name provides an answer to the question "What should we look at?" But the honest answer to that question depends on facts about one's own systems that no name, however clever, can supply. A name can tell you that a danger exists. It cannot tell you where it lives in your house, of if it lives in (or affects) your house at all.
More than a decade after Heartbleed, the practice has settled into something that's mostly less theatrical. Logos have largely receded, surviving mainly in academic research on processor flaws, where "Meltdown" and "Spectre" released theirs into the public domain in 2018, creating something of a tradition in the process.
Names persist, but they have become more like a grammar than a brand. Suffixes such as "-bleed" and "-shell" now tell a reader which earlier catastrophe a new flaw resembles. A name is less often a product launch and more often a shared reference, sometimes coined by the community rather than the discoverer, sometimes adopted by the very vendor whose software is at fault.
As much as I was annoyed by the trend of naming vulnerabilities when it started, this seems to me the right resting place, and it suggests what naming was always for. A name is a promise of continuity. It says: this thing will be spoken of again, it has a history, we will remember it. The CVE number is a promise only to the archive. The name is a promise to people, which is why it can move people, and why it can also deceive them.
In hindsight, the deepest effect of naming vulnerabilities on the industry was not more cynical marketing or a weird pivot towards the attention economy prevalent in social media circles. It was that it forced a technical culture that preferred to speak in numbers to confront the fact that it is also a culture of meaning. Security is not only about what is true of a system. It is about what people believe, fear and attend to, and about who gets to shape those beliefs.