Conducting, Sharing and Remediating, 10 min read

Ready reckoner

Prepare

Scope statement · draft Level-1 DFD · 5–8 people · pre-read (scope, DFD, STRIDE prompts, rating scale) · shared board + register template.

Role Job
Facilitator Method and clock; neutral
Scribe Records live, visibly (not the facilitator)
Architect / dev lead How it works (ask a developer to trace journeys)
Tester How to break and verify
Product owner What matters; risk decisions
Specialists Ops, data, privacy, ML engineer, owners of agent-changeable systems

90-minute agenda

Open 10 (goal, scope, rules: no stupid ideas · threats not blame · park deep dives · write decisions) → Model 20 → Find 40 (boundary by boundary, STRIDE) → Rate & assign 15 → Close 5. Record out within 24 h.

Techniques: silent brainstorming · round robin · persona non grata · cards (EoP, LINDDUN GO, AI decks) · assume breach. Failures: lecture · expert debate (park it) · silence · "can't happen, we use X" · everything 5/5 · vague threats · defensiveness · no decisions. AI sessions: ML/data engineer · lifecycle DFD · untrusted-content entry points · tool/permission table · "if an attacker controlled that text…" · OWASP list check.

Threat statement

[Actor] with [prerequisites] can [action], leading to [impact], affecting [asset]. + weakness · STRIDE · location · existing controls.

Report (6 parts)

Executive summary · scope & assumptions · system model · threat register · mitigations & owners · residual risk & acceptance. Version it with the system; classify it as restricted.

Audience Give them
Developers / testers Tickets, acceptance criteria, tests
Product owner Ranked list, effort, options
Executives One page: business exposure + decisions
Audit / risk Full report + traceability + acceptances
Partners / customers Method and control themes; details only under NDA

Remediate

Response Means
Eliminate (ask first) Remove feature/data/capability
Mitigate Add controls
Transfer Contract/insurance (reputation stays)
Accept Named approver, reason, compensating controls, expiry
  • Prevent + detect + respond, in depth; enforce on the server, not UI or prompt.
  • Ticket: threat ID + statement + control + acceptance criteria + test; owner = a person; SLA (e.g. critical before release · high 30 d · medium 90 d · low next change).
  • Verified = evidence the threat is blocked (attacker-side test, pen-test ref, drill record).
  • Close the loop: update model · threats introduced by fixes · retrospective on misses · re-model on triggers.