VERIS, the Vocabulary for Event Recording and Incident Sharing, is the record format behind Verizon’s Data Breach Investigations Report and the VERIS Community Database (VCDB).[1] It was built to make incidents comparable, and it does that well. Its A4 model records an incident as Actor, Action, Asset and Attribute: who acted, what they did, what was hit and what was lost.
A defender needs one more answer: which weakness was exploited, and therefore which control would have stopped it. VERIS has no field for that. The Action axis comes closest, and it carries everything at once. To measure what that costs, I mapped every VERIS 1.4.2 value to the ten TLCTC clusters and ran the mapping over all 10,047 records of the VCDB.[4][5] This post shows where VERIS breaks, what the data says, and why a cause axis is the right repair.
The weak point: one axis for everything
VERIS’s Action axis holds causes (SQL injection), payload capabilities (command and control), channels (the vectors), properties of a step (Evade Defenses), and events that are not attacks at all (Error, Environmental). The words are not defined tightly enough to carry a cause, and the same word often names several things.
| VERIS term | What it mixes | TLCTC reading |
|---|---|---|
| Hacking | Five different weaknesses in one category: exploited flaws, stolen credentials, brute force, DoS, abuse of functionality. | #1, #2/#3, #4, #6, separated by R-ROLE, R-CRED and R-SPECIFIC |
| Exploit vuln | No record of whether the flawed component served the attacker or processed attacker content. | #2 server role or #3 client role (R-ROLE) |
| Malware varieties (Backdoor, C2, Ransomware, Capture stored data …) | Each names a feature of a payload. Three ticked varieties read as three actions, but there was one execution. | One #7 step at the moment foreign code runs (R-EXEC); the delivery is a separate step |
| Backdoor | Three meanings: a hacking result, a malware capability and a vector. | #1 if built from designed functions, #7 if it is an implant |
| Misuse, Privilege abuse | An insider using resources against their purpose, filed as a threat action. | Abuse of Rights when the actor acts within a genuine grant: operational risk, no cluster (R-SCOPE) |
| Error, Environmental | Misdelivery, lost devices and fire recorded as actions beside attacks. | Error in Use and Failure: real incidents on other rows of the cause-side partition, no cluster |
| Partner vector | A partner that relayed an attack and a partner whose artefact the victim’s system trusted are the same value. | Transit, or #10 at the Trust Acceptance Event when the artefact was subverted (R-SUPPLY) |
| Availability varieties | Data state (Loss, Obscuration) sits next to service state (Interruption, Degradation). | [DRE: Av] data gone, [DRE: Ac] data present but unusable |
The second gap is order. A VERIS record that ticks Social, Hacking and Malware says that all three happened, not which one enabled which. The VERIS maintainers know this: their repository names sequencing as the main feature planned for VERIS 2.0, and the issue “Add Sequencing to VERIS” has been open since June 2016.[3]
What 10,047 records show
The mapping covers every VERIS 1.4.2 action variety, vector and result and every attribute value: 338 entries, each with the rule that decides it and a one-sentence rationale. A deterministic classifier applied it to the VCDB snapshot of 4 August 2026. Six findings stand out.
1. Four in ten records carry no threat at all
4,020 records (40.0%) contain only events that are not attacks: 2,578 errors and environmental events (25.7%) and 1,413 insiders acting within an entitlement they genuinely held (14.1%). These are real incidents, and most of them carry a confirmed data risk event. But no generic vulnerability was exploited, and the controls that prevent them are process design, segregation of duties and purpose auditing, not cyber controls. VERIS keeps them on the same axis as SQL injection.
2. One threat record in three needs a rule VERIS does not record
Of the 4,070 records that do carry a threat, 1,911 (47.0%) resolve to clusters directly. 1,311 (32.2%) cannot be resolved without a rule VERIS does not encode: Exploit vuln in 1,034 records, which is #2 or #3 depending on the role of the flawed component, and Backdoor in 747, which is #1 or #7 depending on whether foreign code ran.
3. One in five has lost the step its cluster requires
848 threat records (20.8%) name a cluster whose rules require a step the record does not contain: 425 records of malware without the step that delivered it, 212 of credential use without the step that obtained the credential, and 211 of social engineering without the technical step that followed. The record says what happened last and loses how it started.
4. Ransomware is recorded without its encryption
Of 1,084 ransomware records, 772 (71.2%) carry no availability attribute at all; since 2020 it is 754 of 873 (86.4%). Most of them do carry a confidentiality disclosure. Whether the data is gone (Av) or present but unusable (Ac) decides between restoring from backup and negotiating for a key. VERIS can express the difference; the records mostly do not.
5. Recent years are dominated by one template
747 of the 1,317 records since 2020 (56.7%) share one five-value signature: Backdoor and Exploit vuln as hacking, three malware varieties ending in Ransomware. It is a coding template for ransomware incidents whose initial access was not reported. The honest TLCTC rendering is ? → #7 + [DRE: C, Ac]: an unresolved first step, malware at execution, and the outcomes.
6. Half of all hacking is “Unknown”
1,695 of 3,368 hacking records give the variety as Unknown. In TLCTC notation these are ?: something happened, no cluster can be defended, and the record is excluded from cluster statistics instead of being counted under a catch-all (R-UNRES-3).
As a cross-check, VERIS’s own VERIS → ATT&CK crosswalk combined with the TLCTC ATT&CK mapping gives a second route to the clusters. Where both routes reach a record, they agree on 2,122 of 2,429 records (87.4%).
VCDB is a sample of publicly reported incidents, with older years and healthcare over-represented. None of these numbers is a population estimate. They show what a VERIS record can and cannot say about cause, which is a property of the format, not of the sample.
Why a cause axis is the right repair
Each of the findings above is answered by a rule TLCTC already has. That is the point of a cause-oriented taxonomy: it defines its categories by the generic vulnerability an attack step exploits, so a category can be tested for membership and two analysts reach the same answer.
- One step, one cluster. Every step exploits exactly one generic vulnerability (Axiom VI). The Hacking grab-bag splits into the clusters it contains, and three malware varieties become one #7 step with three features.
- Outcomes are not causes. What was lost is a data risk event, never a cluster (Axiom III): C, I (refined to Ii, incorrect, or If, misattributed) and A (refined to Av or Ac).
- Not every incident is an attack. Three questions, asked in order, place every event on one row of the cause-side partition: is there an actor, did the actor intend the outcome, did an accountable grantor entitle the action (R-SCOPE)? Errors, failures and in-grant misuse keep their data risk event and leave the threat axis.
- The actor does not classify. Who acted is metadata (Axiom IV). The entitlement question is about the action, not the person.
- Rules decide what VERIS leaves open. R-ROLE for exploited flaws, R-EXEC for foreign code, R-CRED for credentials (acquisition and use are separate steps), R-SPECIFIC for capacity against defect, R-SUPPLY for supply chain.
- An incident is a path. Ordered steps with an interval between them,
#9 →[Δt=24h] #7 →[Δt=5m] #4: exactly the sequencing VERIS has planned for its next major version. - Consequences form a chain. System risk event, data risk event, business risk event. A business risk event is reached either by a causal edge, which carries an interval, or by a normative edge, where a duty arises, which does not.
One record, two readings
A credential-stuffing incident against a web shop, coded in VERIS:
| VERIS | TLCTC |
|---|---|
| Actor: External | Metadata: does not classify |
| Action: Hacking / Use of stolen creds | #4 →[Δt=seconds] #1 + [DRE: C]: the stolen credentials are applied at the login (#4), then the shop’s own export function is used to pull customer data (#1; stolen credentials never confer a grant) |
| Asset: Server / Web application | Where: @Org, server role |
| Attribute: Confidentiality / Personal data | [DRE: C] on the step that produced it; a notification duty under GDPR Art. 33 arises by a normative edge. Its 72-hour deadline bounds the response; it is not an interval of the attack. |
The credentials came from another company’s breach. That happened in the other company’s sphere, so it is not a step in this path, and it is not #10: the shop accepted no subverted trust artefact. The VERIS record cannot express any of this. It names the credential use and stops.
The newest values make the point again
VERIS 1.4.2, released in October 2026, adds “Agentic system” as an actor variety, “Misalignment” as a motive, and “Inadequate containment” as an error: a failure to enforce boundaries around an agent.[1] That is the actor-axis approach applied to AI: the new values say who acted and why. TLCTC asks what was exploited. An agent acting within its grant is not an attack. An agent that breaks out because nobody enforced its boundaries is an Error in Use, with no cluster. An attacker who steers the agent commits an attack step, classified by the weakness it exploits. The actor stays the same in all three cases; the cause does not.
No VERIS rewrite is needed to start
The mapping is published in VERIS’s own format: a CSV in the layout of the VERIS → ATT&CK crosswalk that already sits in the VERIS repository, proposed there as a pull request.[6] Joining it to existing records gives, today, a separation of threats from non-threats, a cluster for nearly half of the threat records with honest bounds for the rest, and Av/Ac wherever the coder filled in availability.
Three things only VERIS can add:
- A role on Exploit vuln (server or client). One enumeration would resolve 1,034 records.
- An entitlement flag on Misuse (within the grant or past it). It would separate Abuse of Rights from attacks in the 1,789 records that carry a misuse action.
- Sequencing. An ordered list of steps, one cluster each, an interval between them and the outcomes on the step that produced them. TLCTC’s attack-path schema is one tested shape for it.
The end state keeps the four letters and repairs what they mean: Actor as metadata, Action as a sequence of cluster steps with intervals, Asset as responsibility sphere and topology, Attribute as the chain from system risk event to data and business risk events.
Bottom line
VERIS asked the right question: how do we record incidents so they can be compared? It answers who, what and what was lost. It cannot answer why, because nothing in it is defined by the weakness exploited. The VCDB shows the cost in numbers: four in ten records that are not threats at all, one in three that cannot be classified without a rule the format lacks, one in five that has lost its first step, and ransomware without its encryption.
A cause axis does not replace VERIS. It gives VERIS the one dimension that turns incident counts into control decisions, and the crosswalk to use it already exists.
Sources
- [1] vz-risk/veris, VERIS 1.4.2 (released 14 October 2026; changelog: issues #508, #510, #512), CC BY-SA 4.0.
- [2] vz-risk/VCDB, joined dataset at commit 230cf22 (4 August 2026), 10,047 records.
- [3] VERIS issue #127, “Add Sequencing to VERIS”, open since June 2016.
- [4] TLCTC
mappings/veris/: the mapping (338 entries), the classifier and the study script. - [5] What a Cause Axis Recovers from VERIS: a TLCTC reading of 10,047 VCDB records (study report); current tables in
study/results.md. - [6] VERIS pull request #507, “Add VERIS 1.4.2 → TLCTC v2.7 mapping”.
- [7] TLCTC v2.7 core paper: Axioms III, IV, VI; R-SCOPE and the cause-side partition; R-ROLE, R-EXEC, R-CRED, R-SPECIFIC, R-SUPPLY; data risk events; the consequence chain.
Related: 2025 DBIR Analysis Through the TLCTC Lens · The Cyber Bow-Tie · RSIT Records the State. TLCTC Names the Cause.
TLCTC Framework · v2.7 · CC BY 4.0