Swarmchasing Report a swarm

Field Manual

Field Manual sections

Every section of the Swarmchasing Field Manual (SC-FM) by its id. Cite a section as SC-FM 1.1 §FM-3.2. Search matches ids, titles and the sections' text; it runs in this page.

1. Scope and definitions

  1. FM-1.1 Scope
  2. FM-1.2 Units of observation
  3. FM-1.3 Groupings
  4. FM-1.4 Casework terms
  5. FM-1.5 Evidence tiers
  6. FM-1.6 Evidence classes in a claim

2. Principles

  1. FM-2.1 Passive only
  2. FM-2.2 Evidence over narrative
  3. FM-2.3 Calibrated attribution
  4. FM-2.4 Admin gate
  5. FM-2.5 Publish what informs, withhold what enables
  6. FM-2.6 Correct in public
  7. FM-2.7 Minimise sensitive data
  8. FM-2.8 Proportionality

3. Sources and collection

  1. FM-3.1 What counts as a public record
  2. FM-3.2 Passivity test
  3. FM-3.3 Source rationale
  4. FM-3.4 Terms of use
  5. FM-3.5 Politeness
  6. FM-3.6 Known source biases
  7. FM-3.7 Redaction at ingest
  8. FM-3.8 Retention

4. Indicators

  1. FM-4.1 Grading records
  2. FM-4.2 Rule anatomy
  3. FM-4.3 The SC-T tactic taxonomy
  4. FM-4.4 Default tactics by rule kind
  5. FM-4.5 Artifact families
  6. FM-4.6 Cross-walk to MITRE ATLAS and ATT&CK
  7. FM-4.7 Rule-writing standard
  8. FM-4.8 Rule versions and pinning
  9. FM-4.9 Benchmark targets and declared testing
  10. FM-4.10 Interaction signals
  11. FM-4.11 Independence groups and the context cap

5. Linking

  1. FM-5.1 Hard and soft links
  2. FM-5.2 Shared link features
  3. FM-5.3 Close timing on the same target
  4. FM-5.4 Probe chains
  5. FM-5.5 Runs
  6. FM-5.6 Cluster hygiene
  7. FM-5.7 Soft links
  8. FM-5.8 Candidate fleets, merge and split
  9. FM-5.9 Escalation sequences
  10. FM-5.10 Reruns
  11. FM-5.11 Provenance and planted records
  12. FM-5.12 Interaction edges
  13. FM-5.13 Cluster lifecycle and identifiers
  14. FM-5.14 Scaffold fingerprints
  15. FM-5.15 Coordinated accounts on public platforms
  16. FM-5.16 Co-targets
  17. FM-5.17 Weighing soft links
  18. FM-5.18 Fleets across windows
  19. FM-5.19 Ordered co-targets
  20. FM-5.20 Scaffold lineage
  21. FM-5.21 Prompt echo
  22. FM-5.22 Co-groups
  23. FM-5.23 Identifier generators
  24. FM-5.24 Playbook alignment
  25. FM-5.25 Campaigns

6. Timing and concurrency

  1. FM-6.1 Timestamp kinds
  2. FM-6.2 Cadence, bursts and pauses
  3. FM-6.3 Concurrency and fan-out
  4. FM-6.4 Rhythm
  5. FM-6.5 Clock offset from date tags
  6. FM-6.6 Stale dates
  7. FM-6.7 Shared pauses
  8. FM-6.8 Stating timing claims
  9. FM-6.9 Co-activity
  10. FM-6.10 Onset groups

7. Attribution

  1. FM-7.1 Four layers
  2. FM-7.2 Developer and operator ladder
  3. FM-7.3 Hosting and runtime layer
  4. FM-7.4 Harness and framework layer
  5. FM-7.5 Model family layer
  6. FM-7.6 Claim tags
  7. FM-7.7 Attribution labels
  8. FM-7.8 The naming gate
  9. FM-7.9 False-flag reasoning
  10. FM-7.10 Agent likelihood
  11. FM-7.11 Declared agent identity
  12. FM-7.12 Competing hypotheses
  13. FM-7.13 Published address ranges

8. Confidence language

  1. FM-8.1 Likelihood table
  2. FM-8.2 Confidence levels
  3. FM-8.3 Usage rules

9. Evaluation

  1. FM-9.1 Three kinds of evaluation
  2. FM-9.2 Regression set
  3. FM-9.3 Frozen held-out split
  4. FM-9.4 Prospective evaluation
  5. FM-9.5 Detection metrics
  6. FM-9.6 Cluster metrics
  7. FM-9.7 Attribution calibration
  8. FM-9.8 Inter-rater agreement
  9. FM-9.9 What to publish with a number
  10. FM-9.10 Benchmark publication rules
  11. FM-9.11 Agent-grade evaluation
  12. FM-9.12 Calibrated public language
  13. FM-9.13 Negative control and benign baseline

10. Disclosure

  1. FM-10.1 Parties
  2. FM-10.2 Lead times
  3. FM-10.3 Relation to vulnerability disclosure standards
  4. FM-10.4 Contacts and channels
  5. FM-10.5 Right of reply
  6. FM-10.6 Authorised testing
  7. FM-10.7 Operator-supplied logs
  8. FM-10.8 The admin gate and the case record
  9. FM-10.9 Self-review and second review

11. Notice standard

  1. FM-11.1 Purpose
  2. FM-11.2 Structure
  3. FM-11.3 Status lifecycle
  4. FM-11.4 What a notice withholds
  5. FM-11.5 Lint
  6. FM-11.6 Corrections and retractions
  7. FM-11.7 Permanence

12. Interoperability

  1. FM-12.1 STIX 2.1 export profile
  2. FM-12.2 The x-sc-agent-run extension
  3. FM-12.3 AI incident databases
  4. FM-12.4 Crosswalk to agent-security frameworks
  5. FM-12.5 Regulatory incident reports

13. Case study template

  1. FM-13.1 Template
  2. FM-13.2 Worked example (synthetic)
  3. FM-13.3 Reading a published write-up with the template
  4. FM-13.4 Case study: urlquery.net agent activity
  5. FM-13.5 Case study: US and Canadian government targets
  6. FM-13.6 Case study: an agent fleet reading a map service

14. Known limits and traps

  1. FM-14.1 Self-identification is forgeable
  2. FM-14.2 An ASN names a host
  3. FM-14.3 A scan service's ASN is the target's
  4. FM-14.4 Capture time is not request time
  5. FM-14.5 Tags get copied after publication
  6. FM-14.6 Common features cause giant clusters
  7. FM-14.7 Humans look the same
  8. FM-14.8 Model residue comes from people too
  9. FM-14.9 Stacking across a cluster
  10. FM-14.10 Malformed URLs make bogus targets
  11. FM-14.11 A record is not a success
  12. FM-14.12 Absence is not evidence
  13. FM-14.13 Display truncation looks like model truncation
  14. FM-14.14 Decoding is heuristic
  15. FM-14.15 Date and rhythm assumptions
  16. FM-14.16 Cron jobs share pauses
  17. FM-14.17 Model-family claims are fragile
  18. FM-14.18 Redaction hashes are not encryption
  19. FM-14.19 Rules change
  20. FM-14.20 Public records can be planted
  21. FM-14.21 A tier is not misalignment
  22. FM-14.22 Correlated rules stack

15. Misalignment and scope

  1. FM-15.1 Four questions
  2. FM-15.2 Task class and target class
  3. FM-15.3 Scope tests
  4. FM-15.4 Stating a scope finding
  5. FM-15.5 From finding to incident

16. Relation to other methods

  1. FM-16.1 Published incident investigations
  2. FM-16.2 Provider threat reports
  3. FM-16.3 Analytic and coordination frameworks
  4. FM-16.4 Research on agent coordination and fingerprints
  5. FM-16.5 Standards work on agents