Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 20 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,26 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

## [Unreleased]

### Added

- Operations dashboard alert-posture counters and severity/status breakdowns
now open the corresponding filtered alert grid. Critical-open links use the
canonical unresolved-status predicate; the 24-hour view uses alert creation
time. Tenant scope continues to come from the authenticated principal.

### Fixed

- Alert drill-down filters now stay in sync with the URL when changed or
cleared, so reloading or restoring browser history does not revive removed
filters. Explicit drill-downs take precedence over default saved views while
preserving manual preset selection. Empty-state guidance
includes unresolved-only and creation-time filters, and posture bar links
have a visible keyboard-focus indicator. Regression tests preserve link
accessibility attributes, accept SQLAlchemy's expanding collection binds,
and verify the alerts client's HTTP query parameter names.

## [23.5.3] - 2026-11-19

### Fixed

- **Checkpoint cursors order and compare the instant a timestamp names, not
Expand Down
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@

**An open-source, self-hostable AI Security Operations Center.** It ingests your security telemetry, detects and correlates threats, investigates them with AI agents whose reasoning is fully auditable, and proposes responses a human approves.

[![License: MIT](https://img.shields.io/badge/License-MIT-22c55e.svg?style=flat-square)](https://opensource.org/licenses/MIT) [![Version](https://img.shields.io/badge/version-23.5.2-f59e0b?style=flat-square)](CHANGELOG.md) [![CI](https://img.shields.io/github/actions/workflow/status/beenuar/AiSOC/ci.yml?branch=main&label=CI&style=flat-square)](https://github.com/beenuar/AiSOC/actions/workflows/ci.yml)
[![License: MIT](https://img.shields.io/badge/License-MIT-22c55e.svg?style=flat-square)](https://opensource.org/licenses/MIT) [![Version](https://img.shields.io/badge/version-23.5.3-f59e0b?style=flat-square)](CHANGELOG.md) [![CI](https://img.shields.io/github/actions/workflow/status/beenuar/AiSOC/ci.yml?branch=main&label=CI&style=flat-square)](https://github.com/beenuar/AiSOC/actions/workflows/ci.yml)
[![CodeQL](https://img.shields.io/github/actions/workflow/status/beenuar/AiSOC/codeql.yml?branch=main&label=CodeQL&style=flat-square)](https://github.com/beenuar/AiSOC/actions/workflows/codeql.yml) [![OpenSSF Scorecard](https://api.securityscorecards.dev/projects/github.com/beenuar/AiSOC/badge)](https://securityscorecards.dev/viewer/?uri=github.com/beenuar/AiSOC) [![Technical Guide](https://img.shields.io/badge/Technical%20Guide-28%20page%20PDF-dc2626?style=flat-square)](https://raw.githubusercontent.com/beenuar/AiSOC/main/apps/web/public/papers/aisoc-technical-guide.pdf)

**[Technical Guide (PDF)](https://raw.githubusercontent.com/beenuar/AiSOC/main/apps/web/public/papers/aisoc-technical-guide.pdf)** · [Docs](https://beenuar.github.io/AiSOC/) · [Architecture](docs/architecture/README.md) · [What actually works](docs/audit/REPOSITORY_REALITY.md) · [Discussions](https://github.com/beenuar/AiSOC/discussions)
Expand Down
20 changes: 18 additions & 2 deletions RELEASES.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,9 @@

This file mirrors what used to live in the "What's new" section of [`README.md`](README.md). The complete, machine-readable inventory (with file paths, env-var diffs, and per-release test counts) lives in [`CHANGELOG.md`](CHANGELOG.md).

> **TL;DR for first-time visitors:** AiSOC is on `v23.5.2`, released 2026-11-19. **The patch that undoes a silent revert the previous two tags shipped.** A stale-base push had serialized an old tree over `main`: the #1933 merge (an unrelated cases fix) quietly rolled back the #1904 security work, so v23.5.0 and v23.5.1 shipped a **fail-open** threat-actor attribution enforcer (unset token → warn once and allow every caller) while both releases' notes described 503 — and managed-tenant provisioning once again published the known-password demo administrator into every newly promoted tenant. Both are fixed here, each the better way: [@knowlet](https://github.com/knowlet)'s #1966 removes the demo identity from provisioning **structurally** (managed tenants get labelled starter content, never a login; `_ensure_user` itself refuses any tenant but the canonical demo tenant; migration 163 retires demo admins already published into managed tenants — deactivated, sessions revoked, hash replaced, role demoted; a live negative control re-weakens both layers and asserts the suite fails), and #1967 restores the fail-closed enforcer (unset token → 503 unless `AISOC_THREATINTEL_ALLOW_UNAUTHENTICATED=1` is set deliberately), its tests, the deleted compose token gate — now one of **two** detectors for that contract — and Dependabot #1928's reverted realtime OpenTelemetry bump. Every merge in the window was audited for the revert signature: exactly two bad merges, both now remediated, and the push tooling that caused them refuses stale worktrees outright. The withdrawn `picomatch` advisory (GHSA-fc5q-39xc-vjpf, retracted upstream 2026-11-18) was dismissed as inaccurate, taking the Dependabot queue to zero. **2 defects were fixed in this release** (the fail-open enforcer and the provisioning identity, counted as shipped-tag defects; the OTel re-apply is a dependency restoration). **No breaking changes**, with one behavioural note: migration 163 retires `demo@example.com` admin rows in managed tenants — the canonical demo tenant is untouched, and an operator who deliberately wants the demo identity re-runs the demo seeder against that tenant. **Neither `stable` channel moves.** The chart publishes as **`7.24.2`** — an `appVersion`-only move carrying v23.5.2. Everything below describes v23.5.1 and earlier, which this release does not change.
> **TL;DR for first-time visitors:** AiSOC is on `v23.5.3`, released 2026-11-19. **The connector data-completeness patch, community end to end: both defects reported with failing tests by [@pavancharak](https://github.com/pavancharak), both fixed by first-time contributor [@Varshith-Kali](https://github.com/Varshith-Kali), verified against the shipped tag.** The sharper one is in the shared checkpoint machinery every cursored connector uses: `apply_checkpoint` ordered and compared vendor timestamps as **text**, and text order is not time order — `…10:00:00.500Z` is half a second *after* a cursor of `…10:00:00Z` but sorts before it (`.` < `Z`), and a cursor written `…11:00:00+05:00` (06:00Z) drops a row at `…10:00:00Z`, four hours later. Rows newer than the cursor were silently dropped as already read, and the staged cursor could name the wrong row. The cursor now orders and compares the parsed UTC instant with the id breaking ties; a timestamp that does not parse keeps its row (an unparseable time is never a reason to drop data) and falls back to the old text comparison. The second defect: **Huntress, Cisco Secure Endpoint and Sophos Central polled bare clock windows** — and the scheduler drops a run when the previous poll overruns (`max_instances=1`), so the interval behind a skipped run was never requested by any later poll, invisibly, because a skip sets no `last_outage_at`. All three now persist an event-time cursor (Huntress on `sent_at` — delivery, preserving raise-once semantics), overlap one poll interval so a single skipped run is covered before any cursor exists, and skipped runs are logged (`connector.scheduler.poll_skipped`). The overlap re-reads are dropped against the cursor, so the cost is a re-request, never a duplicate or a lost row. **Verified, not asserted:** 449 connector tests pass on the fix; the new instant-ordering suite fails 5 of 8 against the pre-fix tree, pinning the defect it claims to fix. **2 defects fixed in this release** (both in the v23.5.2 tag; the reporter pinned them there). **No breaking changes**, with one observable note: batches from the three converted connectors now leave ordered oldest-first by event time rather than page order. **Neither `stable` channel moves.** The chart publishes as **`7.24.3`** — an `appVersion`-only move carrying v23.5.3. Everything below describes v23.5.2 and earlier, which this release does not change.
>
> **Previously:** AiSOC was on `v23.5.2`, released 2026-11-19. **The patch that undoes a silent revert the previous two tags shipped.** A stale-base push had serialized an old tree over `main`: the #1933 merge (an unrelated cases fix) quietly rolled back the #1904 security work, so v23.5.0 and v23.5.1 shipped a **fail-open** threat-actor attribution enforcer (unset token → warn once and allow every caller) while both releases' notes described 503 — and managed-tenant provisioning once again published the known-password demo administrator into every newly promoted tenant. Both are fixed here, each the better way: [@knowlet](https://github.com/knowlet)'s #1966 removes the demo identity from provisioning **structurally** (managed tenants get labelled starter content, never a login; `_ensure_user` itself refuses any tenant but the canonical demo tenant; migration 163 retires demo admins already published into managed tenants — deactivated, sessions revoked, hash replaced, role demoted; a live negative control re-weakens both layers and asserts the suite fails), and #1967 restores the fail-closed enforcer (unset token → 503 unless `AISOC_THREATINTEL_ALLOW_UNAUTHENTICATED=1` is set deliberately), its tests, the deleted compose token gate — now one of **two** detectors for that contract — and Dependabot #1928's reverted realtime OpenTelemetry bump. Every merge in the window was audited for the revert signature: exactly two bad merges, both now remediated, and the push tooling that caused them refuses stale worktrees outright. The withdrawn `picomatch` advisory (GHSA-fc5q-39xc-vjpf, retracted upstream 2026-11-18) was dismissed as inaccurate, taking the Dependabot queue to zero. **2 defects were fixed in this release** (the fail-open enforcer and the provisioning identity, counted as shipped-tag defects; the OTel re-apply is a dependency restoration). **No breaking changes**, with one behavioural note: migration 163 retires `demo@example.com` admin rows in managed tenants — the canonical demo tenant is untouched, and an operator who deliberately wants the demo identity re-runs the demo seeder against that tenant. **Neither `stable` channel moves.** The chart publishes as **`7.24.2`** — an `appVersion`-only move carrying v23.5.2. Everything below describes v23.5.1 and earlier, which this release does not change.
>
> **Previously:** AiSOC was on `v23.5.1`, released 2026-11-18. **The patch that makes v23.5.0's own release note true.** v23.5.0 said its "sixteenth secret" — the threat-actor attribution service token — was generated and wired to both ends; on the tag it was neither, so a fresh `make up` answered **503 for every `/api/v1/actors/*` call** with every secret apparently in place. The generator (`ensure_env`) never minted `AISOC_THREATINTEL_SERVICE_TOKEN`, and compose delivered it only to the API, not to the `threatintel` container that enforces the fail-closed gate nor the `agents` container that sends it. This release actually generates it (the generated-secret count moves 15 → 16) and forwards it to both ends, pinned by a gate that discovers the token's readers from source and asserts compose delivers to each — so the generator and the wiring cannot drift apart again. Two more fixes ride along: **`GET /cases?status=open` matches the open caseload again** (it string-matched the write-path alias literally and returned zero rows, which left the console's "add to existing case" picker permanently empty), and **the documentation portal builds again** (a generated connector page shipped `<your subdomain>.huntress.io`, which MDX parses as an unclosed tag, failing every GitHub Pages deploy for six days while per-PR CI stayed green). The README screenshots and the walkthrough video were also re-captured against a live v23.5.0 stack — real local-model verdicts, the real KEV feed, a genuine full-profile attack graph, and two new tiles for the entity-risk-hosts and related-cases features — but that is documentation, not a runtime change. **3 defects were fixed in this release**, all found by exercising the shipped tag rather than reading the diff. **No breaking changes.** **Neither `stable` channel moves.** The chart publishes as **`7.24.1`** — an `appVersion`-only move carrying v23.5.1. Everything below describes v23.5.0 and earlier, which this release does not change.
>
Expand Down Expand Up @@ -56,7 +58,21 @@ This file mirrors what used to live in the "What's new" section of [`README.md`]

## What's new

`VERSION` is `23.5.2`. The **v23.5.2** patch (2026-11-19) undoes a silent
`VERSION` is `23.5.3`. The **v23.5.3** patch (2026-11-19) is the connector
data-completeness fix, community end to end: @pavancharak reported two silent
data-loss defects against the shipped v23.5.2 tag with failing tests, and
first-time contributor @Varshith-Kali fixed both. The shared checkpoint cursor
compared vendor timestamps as text — dropping rows newer than the cursor when
the same instant was written with fractional seconds or a UTC offset — and now
orders by the parsed instant with id tie-breaks, keeping any row whose time
does not parse. Huntress, Cisco Secure Endpoint and Sophos Central polled bare
clock windows that lost the interval behind a scheduler-skipped run; all three
now persist an event-time cursor, overlap one interval, and skipped runs are
logged. 449 connector tests pass; the new suite fails 5/8 on the pre-fix tree.
2 defects fixed; no breaking changes; neither `stable` channel moves; chart
`7.24.3`.

`VERSION` was `23.5.2`. The **v23.5.2** patch (2026-11-19) undoes a silent
stale-base revert the previous two tags shipped: the #1933 merge had quietly
rolled back #1904, so v23.5.0/v23.5.1 carried a fail-open threat-actor
attribution enforcer (warn-and-allow when the token is unset) against release
Expand Down
2 changes: 1 addition & 1 deletion VERSION
Original file line number Diff line number Diff line change
@@ -1 +1 @@
23.5.2
23.5.3
12 changes: 12 additions & 0 deletions apps/docs/docs/console/operations-dashboard.md
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,18 @@ Grouped by reason, so the fix is visible from the panel. "Nothing rejected" rend

Total, new in 24h, critical still open, plus severity and disposition distributions.

All three counters and every severity/status breakdown entry are links to
the alert-centric grid at `/alerts?view=alerts`, with the relevant filters
preselected. `Critical open` combines `severity=critical` with the same
canonical unresolved-status predicate used by the statistics endpoint;
`New (24h)` filters on alert `created_at`, not the event timestamp.
The drill-down's extra filters appear as removable chips above the grid.
Changing or clearing filters updates the URL, so reloading keeps the current
selection and stays in the alert grid even after the final filter is removed.
Explicit drill-down links take precedence over a default saved view; a saved
view can still be selected manually from the toolbar.
Tenant scope still comes from the logged-in principal, never URL parameters.

This is a **snapshot, not a trend**, and the panel says so. No endpoint in the tree buckets disposition over time, so there is no series to draw. Period-over-period movement is published by `/metrics/funnel` and rendered by the funnel strip at the top of the page; nothing here duplicates it with an arrow it cannot source.

### Detection coverage
Expand Down
4 changes: 2 additions & 2 deletions apps/docs/static/data/scoreboard.json
Original file line number Diff line number Diff line change
Expand Up @@ -4,8 +4,8 @@
"rows": [
{
"date": "2026-11-19",
"agent_version": "v23.5.2",
"commit_sha": "308280d6",
"agent_version": "v23.5.3",
"commit_sha": "420c707f",
"substrate": true,
"eval_mode": "substrate-only",
"mitre_accuracy": 0.97,
Expand Down
4 changes: 2 additions & 2 deletions apps/docs/static/llms-full.txt
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
<!-- Generated by scripts/generate_llms_txt.py. Do not edit by hand; run the script and commit the result. -->

> AiSOC is an open-source, self-hostable AI Security Operations Center, MIT
> licensed and maintained by the AiSOC community. Version 23.5.2.
> licensed and maintained by the AiSOC community. Version 23.5.3.
> Repository: https://github.com/beenuar/AiSOC. Documentation: https://beenuar.github.io/AiSOC/.

This file exists so an automated reader can answer questions about AiSOC
Expand Down Expand Up @@ -127,7 +127,7 @@ plans/ Historical prototype subtree — reference only, never edited

Every number in this file is generated from the artefact that owns it. If you
need to state one, re-derive it rather than quoting a figure from prose —
including from this file, which is a snapshot of version 23.5.2.
including from this file, which is a snapshot of version 23.5.3.

```
python3 scripts/generate_corpus_stats.py --check # executable + on-disk rules
Expand Down
2 changes: 1 addition & 1 deletion apps/docs/static/llms.txt
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@
> security telemetry from 97 first-party connectors, normalises it to OCSF,
> correlates it into alerts, and runs LLM agents that triage and investigate them.
> Every agent decision is written to an investigation ledger, so the reasoning
> behind a verdict can be replayed step by step. MIT licensed. Current version 23.5.2.
> behind a verdict can be replayed step by step. MIT licensed. Current version 23.5.3.

AiSOC is built and maintained by the AiSOC community. It is a complete SOC
platform rather than a single tool: ingestion, detection, correlation, agent
Expand Down
Loading
Loading