Cybersecurity has a language problem: “threat” means a dozen different things. TLCTC fixes the vocabulary on the cause side with ten non-overlapping clusters, and keeps causes apart from the events and consequences they produce. Each cluster names one generic vulnerability — the root weakness an attack step exploits. The rest follows from the power of causality. This summary takes five minutes; the v2.7 core paper holds the full specification.
The Problem
Cybersecurity has a language problem.
We use “threat” to mean a dozen different things: the attacker, the vulnerability, the malware, the breach, the ransomware demand, the missing control. When a CISO briefs the board, when a SOC analyst writes an incident report, when a vendor sells a solution — they often use the same words for fundamentally different things.
This isn’t just semantics. It means we can’t compare incidents. We can’t aggregate threat intelligence cleanly. We can’t map controls to threats with confidence. And we can’t communicate risk consistently across teams.
The Solution
TLCTC (Top Level Cyber Threat Clusters) is a cause-oriented taxonomy. It classifies every step of an attack by the generic vulnerability it exploits — not by who is attacking, which tool is used, or what the outcome is.
Ten clusters. Non-overlapping: one attack step, one generic vulnerability, one cluster. The same ten for any technology and any sector.
| Cluster | In plain words |
|---|---|
| #1 | Abuse of Functions Turning designed features and functions against their purpose — no code flaw involved. |
| #2 | Exploiting Server Exploiting a flaw in software acting in a server role: it receives requests and input. |
| #3 | Exploiting Client Exploiting a flaw in software acting in a client role: it processes content it fetched. |
| #4 | Identity Theft Presenting stolen or forged credentials to act as someone else. |
| #5 | Man in the Middle Sitting between communicating parties to read, alter or relay their traffic. |
| #6 | Flooding Attack Exhausting finite capacity — bandwidth, compute, connections — so a service stops serving. |
| #7 | Malware Getting foreign executable content to run in the environment. |
| #8 | Physical Attack Physical access to hardware, or its physical properties and signals — with or without contact. |
| #9 | Social Engineering Manipulating people’s trust and psychology so they act against security. |
| #10 | Supply Chain Attack Subverting a trusted third party so that its software, update, hardware or service is accepted as authoritative. |
Plain-language summaries for orientation. The canonical definitions, generic vulnerabilities and boundary tests are on the definitions page; how the ten were derived is a short film (4:48).
Why This Matters
Ransomware isn’t a threat category — it’s an outcome.
“Ransomware attack” describes the result (encrypted files, an extortion demand) and maybe the tool. The attack that led there is a sequence of steps, for example:
#9 →[Δt=24h] #7 →[Δt=5m] #4 →[Δt=15m] #1 → #7 + [DRE: Ac]
A phishing e-mail persuades an employee to open an attachment (#9 Social Engineering). The attachment runs code (#7 Malware) that harvests credentials. Minutes later the attacker logs in with them (#4 Identity Theft), uses legitimate admin tools to push the ransomware out (#1 Abuse of Functions), and the ransomware executes (#7). The data is still there, but unusable — a data risk event of type Ac, inaccessible.
Each step exploits a different generic vulnerability, and each has different controls. The arrows carry the time between steps (Δt): five and fifteen minutes is faster than any human response — only automated detection and containment keep up. Calling the whole thing “a ransomware attack” hides what actually happened, and what could have stopped it.
The Key Insight: Causes vs. Consequences
TLCTC is built on the bow-tie risk model, which keeps causes and consequences strictly apart:
- Left side — causes: how did the attacker get in? The ten clusters live here, and only here.
- Centre — the System Risk Event (SRE): loss of control — the moment a system’s behaviour, privileges, data or trust leave its owner’s control.
- Right side — consequences, in layers: a Data Risk Event (DRE: loss of confidentiality, integrity or availability) and, on top of it, Business Risk Events (BRE): first the business process (customer data exposed, payments down), then compliance, legal and financial consequences.
Not every compromise becomes a data loss, and not every data loss becomes a business event. Each causal transition leaves a window in which the chain can still be broken.
This separation matters because:
- The same attack path can lead to different consequences.
- The same consequence can result from different attack paths.
- Preventive controls act on the cause side and lower the likelihood; detective and reactive controls limit the consequences.
The clusters classify attacks. Technical failures, honest mistakes and the misuse of rights someone was legitimately granted (“abuse of rights”) are operational risk with their own controls — segregation of duties, four-eyes, supervision. Keeping them out of the cyber register keeps the CISO accountable for what security controls can actually change. An insider acting outside their grant, or anyone using stolen credentials, is an attack like any other.
What You Can Do With It
- Compare incidents — express any intrusion as a chain of cluster steps, comparable across technologies and sectors.
- Map controls precisely — ten clusters × the six NIST CSF functions give 60 named control objectives; coverage gaps show up as empty cells. (Populated cells structure the controls; whether they work still needs evidence.)
- Measure what matters — set the attacker’s speed per step (Δt) against your time to detect and to contain, and see where you are structurally too slow.
- Report consistently — loss of control, data events and business events are separate events, so each rulebook’s trigger lands on a defined point in the same chain.
- Communicate clearly — give leadership, risk, operations and engineering one vocabulary that means the same thing to everyone.
A supply chain attack (#10) means a trusted third party was subverted and its artifact — software, an update, hardware, a service — was accepted as authoritative inside your domain. A breach at a vendor that holds your data is third-party risk: the attack happens in their sphere, under whichever clusters apply there, and reaches you as a consequence.
Go Deeper
TLCTC v2.7 includes:
- Ten axioms that constrain interpretation and keep classification consistent
- Seventeen classification rules for the recurring boundary cases
- Attack-path notation with velocity (Δt), domain boundaries and unresolved steps
- The event chain System → Data → Business Risk Events, with a refinement tree for data risk events
- Control integration with the NIST CSF functions and key control indicators (application & governance paper)
- Machine-readable JSON schemas and mappings to MITRE ATT&CK, CWE, NIST CSF 2.0, VERIS and more
TLCTC is designed as a translation layer — connecting risk management, security operations, development and executive communication through a shared, precise language for cyber threats.
Films for Every Audience
The same ten clusters, seen from four desks. Each film is a few minutes long and stands on its own.