Skip to content

fix: improve cross-source lifecycle and identity realism - #356

Merged
DavidJBianco merged 26 commits into
devfrom
codex/eforge-expanded-assess-loops-62-71
Jul 14, 2026
Merged

fix: improve cross-source lifecycle and identity realism#356
DavidJBianco merged 26 commits into
devfrom
codex/eforge-expanded-assess-loops-62-71

Conversation

@DavidJBianco

Copy link
Copy Markdown
Collaborator

Summary

  • add a canonical Linux sudo-session action bundle with grouped observation and ordered authorization/PAM lifecycle evidence
  • keep SSH/RDP and general endpoint FLOW observations inside canonical transport intervals and before successful remote authentication
  • correct Windows Security record-ID reset behavior, Kerberos SPN/account identity, and Event 4688 creator-versus-target security context
  • preserve initiating-process identity for DNS while retaining resolver transport ownership
  • align TLS resumption history, DNS sensor timing, and eCAR process/thread attribution with their owning canonical contracts
  • add focused regression coverage, the expanded-assessment worklog, and the updated assessment-effectiveness dashboard

Why

Blind realism loops 62–72 repeatedly exposed sibling defects caused by truth being reconstructed too late or independently in source renderers. These changes move lifecycle, timing, and identity ownership into action bundles, canonical contexts/state, source observation, and source timing so correlated sources agree by construction.

Impact

The generated evidence now preserves remote transport-before-auth ordering, canonical connection intervals, complete sudo lifecycle observation, source-native Windows identity semantics, DNS query-process attribution, and coherent eCAR PID/TID handling. The assessment artifacts document the remaining higher-level work around durable state, packet accounting, stable sensor timing, and final-window admission.

Validation

  • final default suite: 4,911 passed, 19 skipped
  • repository-wide uv run ruff check .
  • repository-wide uv run ruff format --check .
  • configuration and expanded scenario validation
  • per-loop focused regression tests and rendered hard probes
  • final automated evaluation: 95.839, PASS over 93,351 generated records

This PR intentionally targets dev. Version and changelog release updates remain part of the later devmain release PR.

DavidJBianco and others added 26 commits June 27, 2026 10:52
[codex] Release EvidenceForge 1.6.0
fix: optimize web generation performance
Release EvidenceForge 1.9.0
Release EvidenceForge 1.11.0
Release v1.12.0: update agent skill installation
@DavidJBianco
DavidJBianco marked this pull request as ready for review July 14, 2026 18:36
@DavidJBianco
DavidJBianco merged commit a017742 into dev Jul 14, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant