Skip to content

CVD Behavior Tree

This top-level tree integrates all CVD behavioral requirements. See the Requirements Catalog for the full specification:

  • RMB — Report Management behavioral requirements
  • EMB — Embargo Management behavioral requirements
  • CSB — Case State behavioral requirements

We begin at the root node of the CVD Behavior Tree shown in the figure below. The root node is a simple loop that continues until an interrupt condition is met, representing the idea that the CVD practice is meant to be continuous. In other words, we are intentionally not specifying the interrupt condition.

Legacy design diagram

The diagram on this page reflects the simulator-era design from vultron/bt/ and may not match the current implementation in vultron/core/behaviors/. See the Behaviors Reference for auto-generated documentation from the current implementation.

---
title: CVD Behavior Tree
---
flowchart LR
    root["#8634;"]
    seq["→"]
    discover_vulnerability["Discover Vulnerability"]
    receive_messages["Receive Messages"]
    report_management["Report Management"]
    embargo_management["Embargo Management"]
    root --> seq 
    seq -->|A| discover_vulnerability 
    seq -->|B| receive_messages
    seq -->|C| report_management
    seq -->|D| embargo_management

The main sequence is comprised of four main tasks:

  • (A) Discover Vulnerability. Although not all Participants have the ability or motive to discover vulnerabilities, we include it as a task here to call out its importance to the overall CVD process. This task returns Success regardless of whether a vulnerability is found to allow execution to pass to the next task.

  • (B) Receive Messages. All coordination in CVD between Participants is done through the exchange of messages, regardless of how those messages are conveyed, stored, or presented. The receive messages task represents the Participant's response to receiving the various messages defined in the formal protocol. Due to the degree of detail required to cover all the various message types, decomposition of this task node is deferred until later so we can cover the next two items first. Before we proceed, it is sufficient to know that a new report arriving in the receive messages behavior sets \(q^{rm} \in S \xrightarrow{r} R\) and returns Success. Each message is dispatched to one of four process-specific sub-trees:

  • Process RM Messages — handles report-submission, error, and status messages (RS, RE, RV, RI, RA, RD, RC, RK)

  • Process EM Messages — handles all embargo coordination messages (EP, EA, EV, EJ, EC, ET, ER, EE, EK)
  • Process CS Messages — handles case-state change notifications (CP, CX, CA, CV, CF, CD, CE, CK)
  • Process Other Messages — handles general inquiry and acknowledgment messages (GI, GE, GK)

  • (C) Report Management. This task embodies the RM process as integrated into the formal protocol. It delegates to three sub-behaviors plus the main work loop:

  • Report Validation — evaluates credibility and validity; transitions \(q^{rm} \in R \xrightarrow{v} V\) or \(q^{rm} \xrightarrow{i} I\)

  • Report Prioritization — applies triage (e.g., SSVC); transitions \(q^{rm} \in V \xrightarrow{a} A\) or \(q^{rm} \xrightarrow{d} D\)
  • Report Closure — checks closure criteria and transitions \(q^{rm} \xrightarrow{c} C\)
  • Do Work — a parallel node covering the active work performed on Accepted reports (\(q^{rm} \in A\)), including:

    • Deployment — prioritizes and executes fix or mitigation deployment; transitions \(q^{cs} \xrightarrow{\mathbf{D}} D\)
    • Fix Development — Vendor-specific behavior to create a fix; transitions \(q^{cs} \xrightarrow{\mathbf{F}} F\)
    • Reporting to Others — identifies relevant Participants and notifies them (emits RS)
    • Publication — manages preparation and release of advisories, fixes, and exploits; transitions \(q^{cs} \xrightarrow{\mathbf{P}} P\)
    • Monitor Threats — watches for attacks, public exploits, and public reports; may trigger embargo termination
    • CVE ID Assignment — assigns or requests a CVE ID for the vulnerability
    • Acquire Exploit — finds, develops, or purchases a PoC exploit when needed
  • (D) Embargo Management. Similarly, this task represents the EM process as integrated into the formal protocol. It uses three reusable sub-trees:

  • Propose Embargo — selects terms and emits EP (new proposal) or EV (revision), respecting embargo viability conditions

  • Evaluate Embargo — accepts a proposal (emits EA) or issues a counterproposal via Propose Embargo
  • Terminate Embargo — exits an active or proposed embargo when termination criteria are met; transitions \(q^{em} \xrightarrow{t} X\) and emits ET or ER

Behaviors and state changes resulting from changes to the CS model are scattered throughout the other Behavior Trees where relevant.

Section Overview

The diagram below maps the full decomposition of Behavior Trees covered in this section. Each box corresponds to a page in this section.

---
title: CVD Behavior Tree — Section Map
---
flowchart LR
    cvd["CVD BT\n(this page)"]
    vuldisco["Discover Vulnerability"]
    cvd -->|A| vuldisco
    recv["Receive Messages"]
    cvd -->|B| recv
    rm_bt["Report Management"]
    cvd -->|C| rm_bt
    em_bt["Embargo Management"]
    cvd -->|D| em_bt

    proc_rm["Process RM Messages"]
    recv --> proc_rm
    proc_em["Process EM Messages"]
    recv --> proc_em
    proc_cs["Process CS Messages"]
    recv --> proc_cs
    proc_other["Process Other Messages"]
    recv --> proc_other

    rm_val["Report Validation"]
    rm_bt --> rm_val
    rm_pri["Report Prioritization"]
    rm_bt --> rm_pri
    rm_clo["Report Closure"]
    rm_bt --> rm_clo
    do_work["Do Work"]
    rm_bt --> do_work

    deploy["Deployment"]
    do_work --> deploy
    fix_dev["Fix Development"]
    do_work --> fix_dev
    reporting["Reporting to Others"]
    do_work --> reporting
    pub["Publication"]
    do_work --> pub
    mon["Monitor Threats"]
    do_work --> mon
    id_assign["CVE ID Assignment"]
    do_work --> id_assign
    acq_exp["Acquire Exploit"]
    do_work --> acq_exp

    em_prop["Propose Embargo"]
    em_bt --> em_prop
    em_term["Terminate Embargo"]
    em_bt --> em_term
    em_eval["Evaluate Embargo"]
    em_bt --> em_eval