Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
Python agentic-aiai-securityalert-triageblue-teamcybersecuritydetection-engineeringincident-responselanggraphllm-agentsmcpmitre-attackmodel-context-protocolpurple-teamsecurity-operationsself-hostedsiemsoarsocthreat-detectionthreat-intelligence
chore(release): v16.0.1 (#1115)
Beenu Arora056f80a71cf2
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex cc65088da..10c97bf4b 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -7,6 +7,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0## [Unreleased]+## [16.0.1] - 2026-10-04+### Security- **A grant was measured against the caller's static role rather than thediff --git a/README.md b/README.mdindex c986fb151..8593a3aa9 100644--- a/README.md+++ b/README.md@@ -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.-[](https://opensource.org/licenses/MIT) [](CHANGELOG.md) [](https://github.com/beenuar/AiSOC/actions/workflows/ci.yml)+[](https://opensource.org/licenses/MIT) [](CHANGELOG.md) [](https://github.com/beenuar/AiSOC/actions/workflows/ci.yml)[](https://github.com/beenuar/AiSOC/actions/workflows/codeql.yml) [](https://securityscorecards.dev/viewer/?uri=github.com/beenuar/AiSOC) [](https://github.com/beenuar/AiSOC/blob/main/apps/web/public/papers/aisoc-technical-guide.pdf)**[Technical Guide (PDF)](https://github.com/beenuar/AiSOC/blob/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)diff --git a/RELEASES.md b/RELEASES.mdindex a24e70d98..8205a864e 100644--- a/RELEASES.md+++ b/RELEASES.md@@ -2,15 +2,27 @@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 `v16.0.0`, released 2026-10-03. **One breaking change:** the `cases` table is gone and its rows live in `aisoc_cases`. Any SQL of your own against `cases` — a Grafana panel, a scheduled export — now fails with "relation does not exist" rather than returning stale rows, which is the intended behaviour. The theme is a single shape repeated seventeen times: **a mechanism that is complete, tested, and has no caller**, so it passes CI forever while doing nothing. SSO could not sign anyone in — `aisoc_sso_connections` was written by nothing, so every SAML and OIDC callback returned 403 on every deployment, while both handlers were fully tested. MTTA published a confident `0.0` because `alerts.first_seen_at` had no writer. The evidence chain an auditor exports was always `[]`. SLA reporting ran 350 lines of correct arithmetic over an empty table. The console's rule-Approve button could not succeed. The live agent evaluation imported a class that exists only in the historical prototype. And the two case tables never synchronised, so **every case an analyst created was invisible to every case metric** — which is the break above. Measured, not asserted: an agent answering "true positive" to everything scored **1.000** on the detection corpus and now scores **0.727**, because the benign class was derived from `response_class == "monitor"` and those incidents are BloodHound enumeration tagged T1087.002. The first live behavioural run published "unsafe action proposal rate 0.0%", which meant the agent proposed no actions at all; it now reports **not measured** and the verdict flip rate is a real **9.3%**. New: detection lifecycle with rollback and separation of duties, content packs, alert prioritisation, case queues and escalation, legal hold and residency, workload identity and time-boxed elevation, restore paths for Neo4j/Qdrant/Redis, and inbound translation from Splunk SPL, Sentinel KQL and Elastic EQL — on the 2,005 rules bundled here, 1,734 translate and **1,711 of those only partially**. A 22-page technical guide ships as a PDF with real console screenshots. The chart publishes as **`7.3.0`**. Everything below describes v15.1.0 and earlier, which this release does not change.+> **TL;DR for first-time visitors:** AiSOC is on `v16.0.1`, released 2026-10-04. **A security release, no breaking changes.** It closes **GHSA-4gx4-x7gm-4xq8**, a high-severity privilege escalation reported by [HaiND](https://github.com/Haind03): the check that decides whether a caller may *confer* authority read the caller's static role, while the check that admits a caller to the route read its database-resolved permissions. Those are two different answers whenever a tenant uses database-backed RBAC — which is the configuration where it matters, since narrowing an account is what those tables are for. A `tenant_admin` deliberately restricted to `users:write` still carried 28 permissions statically and could grant itself the other 27. **If you use the `roles` and `user_roles` tables to restrict accounts, upgrade.** Deployments with no RBAC rows were never exposed, because the static map was already the correct answer for them. The report named one route; the resolver is shared, so five more had it — authoring a role, re-permissioning one, creating a user, delegating to a child tenant, and minting an API key, that last needing no target user and yielding a durable credential. `scripts/check_role_grant_scope.py` gains a direction that fails CI when a grant is measured against the wrong authority; its previous directions passed on the vulnerable tree because they asked only whether a route *reached* the chokepoint, and all six did. Also raises `serialize-javascript` to `>=7.1.2` and `http-cache-semantics` to `>=4.3.0`. The chart publishes as **`7.3.1`**. Everything below describes v16.0.0 and earlier, which this release does not change.>-> **Previously:** AiSOC was on `v15.1.0`, released 2026-10-03. **No breaking changes** — two routes are added and none removed, so every generated SDK client keeps working. The theme is the gap between a capability existing and a user reaching it: `v15.0.0` closed thirteen security defects sharing the shape *a control that exists, passes its tests and never runs*, and this release applies the same reading to the product. Playbooks could not act at all — `find_matching()` had no production caller, so no playbook had ever run from an alert, and the `approval` step was documented as a durable pause with nowhere to suspend to. A tenant's detection tuning never reached the streaming engine, so a rule turned off in the console kept firing. Every CloudTrail event collapsed onto a single alert, which a one-event pipeline test cannot reveal. Three things are worth reading before upgrading: **first run changed substantially** (`make up` resolves a port conflict instead of refusing to start, measured at 64 seconds from clone to signed-in console on a host with 5432 and 11434 both taken, and a new tenant lands on a setup wizard), **CloudTrail users will see more alerts**, and **the project-maturity table now means something** — `Stable` has a written definition in `docs/audit/MATURITY_DEFINITION.md` and a gate, and the two rows that claimed CI coverage which existed nowhere were fixed rather than relabelled. New: signed, replayable evidence bundles at `GET /api/v1/investigations/{run_id}/bundle`, byte-identical across exports with prompts as digests and an OCSF 1.9.0 mapping checked against the published schema before use; and a copilot that cites every checkable claim to a ledger entry or labels it **uncited**. On the benchmark page, **verdict accuracy is published as *not measured*, with the reason** — every labelled corpus here is entirely malicious by construction, so an agent answering "true positive" to everything would post 100%, and `scripts/score_replay_set.py` refuses such a corpus with a test asserting this tree's own two are refused. The chart publishes as **`7.2.0`**. Everything below describes v15.0.0 and earlier, which this release does not change.+> **Previously:** AiSOC was on `v16.0.0`, released 2026-10-03. **One breaking change:** the `cases` table is gone and its rows live in `aisoc_cases`. Any SQL of your own against `cases` — a Grafana panel, a scheduled export — now fails with "relation does not exist" rather than returning stale rows, which is the intended behaviour. The theme is a single shape repeated seventeen times: **a mechanism that is complete, tested, and has no caller**, so it passes CI forever while doing nothing. SSO could not sign anyone in — `aisoc_sso_connections` was written by nothing, so every SAML and OIDC callback returned 403 on every deployment, while both handlers were fully tested. MTTA published a confident `0.0` because `alerts.first_seen_at` had no writer. The evidence chain an auditor exports was always `[]`. SLA reporting ran 350 lines of correct arithmetic over an empty table. The console's rule-Approve button could not succeed. The live agent evaluation imported a class that exists only in the historical prototype. And the two case tables never synchronised, so **every case an analyst created was invisible to every case metric** — which is the break above. Measured, not asserted: an agent answering "true positive" to everything scored **1.000** on the detection corpus and now scores **0.727**, because the benign class was derived from `response_class == "monitor"` and those incidents are BloodHound enumeration tagged T1087.002. The first live behavioural run published "unsafe action proposal rate 0.0%", which meant the agent proposed no actions at all; it now reports **not measured** and the verdict flip rate is a real **9.3%**. New: detection lifecycle with rollback and separation of duties, content packs, alert prioritisation, case queues and escalation, legal hold and residency, workload identity and time-boxed elevation, restore paths for Neo4j/Qdrant/Redis, and inbound translation from Splunk SPL, Sentinel KQL and Elastic EQL — on the 2,005 rules bundled here, 1,734 translate and **1,711 of those only partially**. A 22-page technical guide ships as a PDF with real console screenshots. The chart publishes as **`7.3.0`**. Everything below describes v15.1.0 and earlier, which this release does not change.+>+> **Before that:** AiSOC was on `v15.1.0`, released 2026-10-03. **No breaking changes** — two routes are added and none removed, so every generated SDK client keeps working. The theme is the gap between a capability existing and a user reaching it: `v15.0.0` closed thirteen security defects sharing the shape *a control that exists, passes its tests and never runs*, and this release applies the same reading to the product. Playbooks could not act at all — `find_matching()` had no production caller, so no playbook had ever run from an alert, and the `approval` step was documented as a durable pause with nowhere to suspend to. A tenant's detection tuning never reached the streaming engine, so a rule turned off in the console kept firing. Every CloudTrail event collapsed onto a single alert, which a one-event pipeline test cannot reveal. Three things are worth reading before upgrading: **first run changed substantially** (`make up` resolves a port conflict instead of refusing to start, measured at 64 seconds from clone to signed-in console on a host with 5432 and 11434 both taken, and a new tenant lands on a setup wizard), **CloudTrail users will see more alerts**, and **the project-maturity table now means something** — `Stable` has a written definition in `docs/audit/MATURITY_DEFINITION.md` and a gate, and the two rows that claimed CI coverage which existed nowhere were fixed rather than relabelled. New: signed, replayable evidence bundles at `GET /api/v1/investigations/{run_id}/bundle`, byte-identical across exports with prompts as digests and an OCSF 1.9.0 mapping checked against the published schema before use; and a copilot that cites every checkable claim to a ledger entry or labels it **uncited**. On the benchmark page, **verdict accuracy is published as *not measured*, with the reason** — every labelled corpus here is entirely malicious by construction, so an agent answering "true positive" to everything would post 100%, and `scripts/score_replay_set.py` refuses such a corpus with a test asserting this tree's own two are refused. The chart publishes as **`7.2.0`**. Everything below describes v15.0.0 and earlier, which this release does not change.---## What's new-`VERSION` is `16.0.0`. The **v16.0.0** release (2026-10-03) carries one+`VERSION` is `16.0.1`. The **v16.0.1** release (2026-10-04) is a security+release closing one high-severity privilege escalation+(**GHSA-4gx4-x7gm-4xq8**): a grant was measured against the caller's static+role rather than the permissions it was admitted on, so an account narrowed+through the database-backed RBAC tables could confer on itself anything its+unnarrowed role carried. Six routes shared the defect, not the one reported.+Upgrade if you use `roles` and `user_roles` to restrict accounts; a+deployment with no RBAC rows was never exposed. The full entry is in+[`CHANGELOG.md`](CHANGELOG.md).++The **v16.0.0** release (2026-10-03) carried onebreaking change: the `cases` table is gone and its rows live in`aisoc_cases`, so SQL of your own against the old name now fails loudlyrather than returning stale rows. It closes seventeen defects sharing onediff --git a/VERSION b/VERSIONindex 946789e61..7141108f1 100644--- a/VERSION+++ b/VERSION@@ -1 +1 @@-16.0.0+16.0.1diff --git a/apps/docs/docs/operations/upgrades.md b/apps/docs/docs/operations/upgrades.mdindex 551188846..ab39c7a1b 100644--- a/apps/docs/docs/operations/upgrades.md+++ b/apps/docs/docs/operations/upgrades.md@@ -142,6 +142,7 @@ this table says what an operator has to **do**.| v14, **v15.0.0** | `/api/v1/shifts` is removed and `/api/v1/threatintel/stix/*` reads answer 404 outside demo mode. `make up` starts a production-class stack and `AISOC_DEV_MODE` no longer defaults to on. | Delete any caller of those routes; neither served data anyone entered. Set `ENVIRONMENT=development` explicitly if you were relying on the dev auth bypass, and expect previously silent warnings to become boot refusals. || v15.0.0, **v15.1.0** | Nothing breaks — two routes are added and none removed. Two behaviours change noticeably all the same: **CloudTrail deployments will see more alerts**, because every event previously collapsed onto a single one; and alert-triggered playbooks now actually run, where `find_matching()` had no production caller before. | No action required to upgrade. Expect the CloudTrail alert volume you should have been getting all along, and review which playbooks are enabled before upgrading, since a playbook that never fired will now fire — in preview by default, with three opt-ins needed before it can act. || v15.1.0, **v16.0.0** | **The `cases` table is gone.** Migration 083 moves its rows into `aisoc_cases`, repoints the child foreign keys, and renames the old table to `cases_pre_consolidation` so nothing is lost. The two tables never synchronised: the console wrote one and every case metric read the other, so MTTR and the case counts were blind to every case an analyst created. Separately, SAML and OIDC sign-in now works — it returned 403 on every deployment because the connection table had no writer. | **Check your own SQL.** Anything querying `cases` — a Grafana panel, a scheduled export, a report — must read `aisoc_cases`. It will fail loudly with "relation does not exist" rather than returning stale rows, which is deliberate. The archive table is readable if you need to compare. Expect case metrics to change the moment you upgrade: they were under-reporting, not over-reporting. If you use SSO, create a connection at `POST /api/v1/sso-connections` before pointing users at it. |+| v16.0.0, **v16.0.1** | **Nothing breaks.** A security release closing a high-severity privilege escalation (GHSA-4gx4-x7gm-4xq8): the check deciding whether a caller may *confer* authority read the caller's static role, while the check admitting it to the route read its database-resolved permissions. Where a tenant uses the `roles` and `user_roles` tables to narrow an account, those are different answers, and the narrowed account could grant itself anything its unnarrowed role carried. | **Upgrade if you restrict accounts through database-backed RBAC.** A deployment with no RBAC rows was never exposed — the static map was already the correct answer for it. After upgrading, a grant is refused when it exceeds what the caller actually holds, so an operator who had been relying on the wider static role to assign roles will now get HTTP 403; grant that operator the permissions through RBAC rather than widening its static role. |Upgrading more than one major at a time is supported but untested as a singlestep. Do them one at a time, running migrations between each, so that adiff --git a/apps/docs/static/data/scoreboard.json b/apps/docs/static/data/scoreboard.jsonindex 9544b49e5..ca57661fd 100644--- a/apps/docs/static/data/scoreboard.json+++ b/apps/docs/static/data/scoreboard.json@@ -4,8 +4,8 @@"rows": [{"date": "2026-10-03",- "agent_version": "v16.0.0",- "commit_sha": "167077e4",+ "agent_version": "v16.0.1",+ "commit_sha": "b1c6925a","substrate": true,"eval_mode": "substrate-only","mitre_accuracy": 0.97,diff --git a/apps/docs/static/llms-full.txt b/apps/docs/static/llms-full.txtindex dcac570de..b89aec214 100644--- a/apps/docs/static/llms-full.txt+++ b/apps/docs/static/llms-full.txt@@ -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 16.0.0.+> licensed and maintained by the AiSOC community. Version 16.0.1.> Repository: https://github.com/beenuar/AiSOC. Documentation: https://beenuar.github.io/AiSOC/.This file exists so an automated reader can answer questions about AiSOCdiff --git a/apps/docs/static/llms.txt b/apps/docs/static/llms.txtindex 3de5881fb..5b6a914b1 100644--- a/apps/docs/static/llms.txt+++ b/apps/docs/static/llms.txt@@ -6,7 +6,7 @@> security telemetry from 84 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 16.0.0.+> behind a verdict can be replayed step by step. MIT licensed. Current version 16.0.1.AiSOC is built and maintained by the AiSOC community. It is a complete SOCplatform rather than a single tool: ingestion, detection, correlation, agentdiff --git a/apps/web/package.json b/apps/web/package.jsonindex b09dbcba4..cd271b8a4 100644--- a/apps/web/package.json+++ b/apps/web/package.json@@ -1,6 +1,6 @@{"name": "@aisoc/web",- "version": "16.0.0",+ "version": "16.0.1","private": true,"description": "AiSOC SOC Console - Next.js 14 Frontend","license": "MIT",diff --git a/infra/helm/aisoc/Chart.yaml b/infra/helm/aisoc/Chart.yamlindex 29a697ea4..2b670504b 100644--- a/infra/helm/aisoc/Chart.yaml+++ b/infra/helm/aisoc/Chart.yaml@@ -25,13 +25,13 @@ type: application# the chart refuses grants it used to accept (see the v14.0.0 BREAKING# section), and the chart's major is the only coordinate a consumer pulling# `oci://ghcr.io/beenuar/charts/aisoc` sees before they install it.-version: 7.3.0+version: 7.3.1# Used as the image tag wherever a `tag:` in values.yaml is left empty, so# it has to be a tag the registry actually holds — it read "5.2.0", which# exists for no image, so every pod of a default install would have gone to# ImagePullBackOff. scripts/check_published_images.py resolves it the same# way the templates do and fails when it names nothing.-appVersion: "v16.0.0"+appVersion: "v16.0.1"keywords:- security