Merge pull request #16094 from Security-Onion-Solutions/saltthangs

Saltthangs
This commit is contained in:
Josh Patterson
2026-07-22 08:44:08 -04:00
committed by GitHub
15 changed files with 418 additions and 56 deletions
+1
View File
@@ -133,6 +133,7 @@ if [[ $EXCLUDE_STARTUP_ERRORS == 'Y' ]]; then
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|Cancelling deferred write event maybeFenceReplicas because the event queue is now closed" # Kafka controller log during shutdown/restart
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|Redis may have been restarted" # Redis likely restarted by salt
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|file already closed" # Go logging race condition during container restart
EXCLUDED_ERRORS="$EXCLUDED_ERRORS|relation \"audit_settings\" does not exist" # salt checking for changes before SOC starts
fi
if [[ $EXCLUDE_FALSE_POSITIVE_ERRORS == 'Y' ]]; then
+10
View File
@@ -150,6 +150,16 @@ logrotate:
- extension .log
- dateext
- dateyesterday
/opt/so/log/postgres/*_x_log:
- daily
- rotate 14
- missingok
- copytruncate
- compress
- create
- extension .log
- dateext
- dateyesterday
/opt/so/log/telegraf/*_x_log:
- daily
- rotate 14
+7
View File
@@ -91,6 +91,13 @@ logrotate:
multiline: True
global: True
forcedType: "[]string"
"/opt/so/log/postgres/*_x_log":
description: List of logrotate options for this file.
title: /opt/so/log/postgres/*.log
advanced: True
multiline: True
global: True
forcedType: "[]string"
"/opt/so/log/telegraf/*_x_log":
description: List of logrotate options for this file.
title: /opt/so/log/telegraf/*.log
+18 -25
View File
@@ -880,29 +880,21 @@ recollate_postgres() {
echo ""
}
bootstrap_so_soc_database() {
# init-db.sh is mounted into so-postgres at /docker-entrypoint-initdb.d/init-db.sh
# and runs automatically only on a fresh data directory. Hosts upgrading from
# 3.1.0 already have /nsm/postgres populated, so the so_soc bootstrap block
# added in 3.2 never fires. Re-run the script explicitly; it's idempotent.
echo "Bootstrapping database via init-db.sh."
# The postgres image has no USER directive, so `docker exec` defaults to
# root, and the container env intentionally omits POSTGRES_USER (the upstream
# entrypoint defaults it transiently during first-init only). Recreate both
# so psql inside init-db.sh resolves the connect user correctly.
local exec_cmd="docker exec -u postgres -e POSTGRES_USER=postgres so-postgres bash /docker-entrypoint-initdb.d/init-db.sh"
if ! /usr/sbin/so-postgres-wait; then
FINAL_MESSAGE_QUEUE+=("WARNING: so-postgres was not ready during the 3.2.0 upgrade; the so_soc database may not have been bootstrapped. Re-run manually: $exec_cmd")
scrub_postgres_log_passwords() {
# Purge plaintext passwords a pre-3.2 postgres could log on DDL errors.
local log=/opt/so/log/postgres/postgres.log
[[ -f "$log" ]] || return 0
if ! grep -qai "PASSWORD '" "$log" 2>/dev/null; then
echo "No leaked passwords found in $log."
return 0
fi
if ! $exec_cmd; then
FINAL_MESSAGE_QUEUE+=("WARNING: init-db.sh failed inside so-postgres during the 3.2.0 upgrade; the database may not have been bootstrapped. Re-run manually: $exec_cmd")
return 0
fi
echo "Database bootstrap complete."
echo "Restarting so-soc container to pick up database changes"
docker restart so-soc
echo "Removing leaked password statements from $log."
local tmp
tmp=$(mktemp)
# Rewrite in place (cat >) to keep the inode postgres is writing to.
grep -avi "PASSWORD '" "$log" > "$tmp" 2>/dev/null || true
cat "$tmp" > "$log"
rm -f "$tmp"
}
# Existing grids should keep ILM unless an admin explicitly opts in to DLM.
@@ -989,7 +981,8 @@ post_to_3.2.0() {
# Recollate due to image OS rebase
recollate_postgres
bootstrap_so_soc_database
# SOC database bootstrap is handled by the postgres.enabled highstate.
scrub_postgres_log_passwords
# Generate 9.3.7 elastic agent installers
echo "Regenerating Elastic Agent Installers"
@@ -2005,11 +1998,11 @@ main() {
# Testing that salt-master is up by checking that is it connected to itself
check_saltmaster_status
# update the salt-minion configs here and start the minion
# update the salt-master and salt-minion configs here and start the minion
# since highstate are disabled above, minion start should not trigger a highstate
echo ""
echo "Ensuring salt-minion configs are up-to-date."
salt-call state.apply salt.minion -l info queue=True
echo "Ensuring salt-master and salt-minion configs are up-to-date."
salt-call state.apply salt.master -l info queue=True
echo ""
# ensure the mine is updated and populated before highstates run, following the salt-master restart
+17
View File
@@ -85,6 +85,23 @@ so-postgres:
- x509: postgres_crt
- x509: postgres_key
postgres_wait_ready:
cmd.run:
- name: /usr/sbin/so-postgres-wait
- require:
- docker_container: so-postgres
- file: postgres_sbin
# Reconcile the SOC database (role, grants, so_telegraf) every highstate so a
# partially-initialized cluster self-heals. POSTGRES_USER is injected because
# the container env omits it.
postgres_bootstrap_soc_db:
cmd.run:
- name: docker exec -u postgres -e POSTGRES_USER=postgres so-postgres bash /docker-entrypoint-initdb.d/init-db.sh
- require:
- cmd: postgres_wait_ready
- file: postgresinitdb
delete_so-postgres_so-status.disabled:
file.uncomment:
- name: /opt/so/conf/so-status/so-status.conf
+8 -4
View File
@@ -8,13 +8,17 @@ if [ -z "${SO_POSTGRES_PASS:-}" ] && [ -n "${SO_POSTGRES_PASS_FILE:-}" ] && [ -r
SO_POSTGRES_PASS="$(< "$SO_POSTGRES_PASS_FILE")"
fi
psql -v ON_ERROR_STOP=1 --username "$POSTGRES_USER" --dbname "$POSTGRES_DB" <<-EOSQL
-- Keep the password out of postgres.log if this DDL errors.
SET log_min_error_statement = panic;
-- Idempotent, race-safe upsert: CREATE, falling back to ALTER if the role
-- already exists or is created concurrently.
DO \$\$
BEGIN
IF NOT EXISTS (SELECT FROM pg_catalog.pg_roles WHERE rolname = '${SO_POSTGRES_USER}') THEN
BEGIN
EXECUTE format('CREATE ROLE %I WITH LOGIN PASSWORD %L', '${SO_POSTGRES_USER}', '${SO_POSTGRES_PASS}');
ELSE
EXECUTE format('ALTER ROLE %I WITH PASSWORD %L', '${SO_POSTGRES_USER}', '${SO_POSTGRES_PASS}');
END IF;
EXCEPTION WHEN duplicate_object OR unique_violation THEN
EXECUTE format('ALTER ROLE %I WITH LOGIN PASSWORD %L', '${SO_POSTGRES_USER}', '${SO_POSTGRES_PASS}');
END;
END
\$\$;
GRANT ALL ON SCHEMA public TO "$SO_POSTGRES_USER";
+2 -11
View File
@@ -8,23 +8,14 @@
{% from 'vars/globals.map.jinja' import GLOBALS %}
{% from 'telegraf/map.jinja' import TELEGRAFMERGED %}
{# postgres_wait_ready below requires `docker_container: so-postgres`, which is
declared in postgres.enabled. Include it here so state.apply postgres.telegraf_users
on its own (e.g. from orch.deploy_newnode) still has that ID in scope. Salt
de-duplicates the circular include. #}
{# postgres.enabled declares the so-postgres container and postgres_wait_ready
that the requires below reference. Salt de-duplicates the circular include. #}
include:
- postgres.enabled
{% set TG_OUT = TELEGRAFMERGED.output | upper %}
{% if TG_OUT in ['POSTGRES', 'BOTH'] %}
postgres_wait_ready:
cmd.run:
- name: /usr/sbin/so-postgres-wait
- require:
- docker_container: so-postgres
- file: postgres_sbin
# Ensure the shared Telegraf database exists. init-db.sh only runs on a
# fresh data dir, so hosts upgraded onto an existing /nsm/postgres volume
# would otherwise never get so_telegraf.
@@ -71,6 +71,8 @@ EOSQL
-v role_user="$ROLE_USER" \
-v role_pass="$ROLE_PASS" \
-U postgres -d so_telegraf <<'EOSQL'
-- Keep the password out of postgres.log if this DDL errors.
SET log_min_error_statement = panic;
SELECT format(
CASE WHEN EXISTS (SELECT FROM pg_catalog.pg_roles WHERE rolname = :'role_user')
THEN 'ALTER ROLE %I WITH LOGIN PASSWORD %L'
+3 -1
View File
@@ -94,7 +94,9 @@ salt_master_service:
- file: checkmine_engine
- file: pillarWatch_engine
- file: engines_config
- order: 9002
- require:
- cmd: wait_for_salt_minion_ready
- order: last
{% else %}
+4 -3
View File
@@ -1530,15 +1530,16 @@ soc:
healthTimeoutSeconds: 5
agentic: false
agentMapping:
Orchestrator: sonnet
Hunter: sonnet
Orchestrator: Claude Sonnet
Investigator: Claude Sonnet
Detection Engineer: Claude Sonnet
onionconfig:
saltstackDir: /opt/so/saltstack
bypassEnabled: false
postgres:
host: ""
port: 5432
sslMode: "allow"
sslMode: "require"
database: securityonion
user: ""
password: ""
@@ -14,18 +14,22 @@
CommandLine: process.command_line
CurrentDirectory: process.working_directory
Details: registry.data.strings
Image: process.executable
ImageLoaded: dll.name
ImageLoaded: dll.path
ParentImage: process.parent.executable
ParentName: process.parent.name
ParentProcessGuid: process.parent.entity_id
ProcessGuid: process.entity_id
TargetFilename: file.name
TargetFilename: file.path
TargetObject: registry.path
TargetUserName: user.target.name
User: user.name
community_id: network.community_id
dns_resolved_ip: dns.resolved_ip
QueryName: dns.question.name
dll_basename: dll.name
file_basename: file.name
document_id: soc_id
dst_ip: destination.ip
dst_port: destination.port
@@ -40,10 +44,15 @@ public_ip: network.public_ip
related_hosts: related.hosts
related_ip: related.ip
src_ip: source.ip
dns_query_name: dns.query_name
dns_query_name: dns.question.name
flow_id: log.id.uid
payload: network.data.decoded
rule_category: rule.category
rule_name: rule.name
rule_uuid: rule.uuid
src_port: source.port
# PowerShell channel (ps_script EID 4104 / classic EID 400)
ScriptBlockText: powershell.file.script_block_text
script_block_id: powershell.file.script_block_id
script_block_hash: powershell.file.script_block_hash
EngineVersion: powershell.engine.version
+121 -2
View File
@@ -1,8 +1,9 @@
name: Security Onion - Playbook Pipeline
priority: 97
transformations:
# Route string fields to their lowercase-normalized .caseless subfield so wildcard
# matches are case-insensitive.
# Route to lowercase-normalized .caseless subfields for case-insensitive matching.
# file.path.caseless exists on Defend only (Sysmon file events lack it);
# registry.path / dll.path / file.name have no .caseless on any source.
- id: case_insensitive_string_fields
type: field_name_mapping
mapping:
@@ -10,3 +11,121 @@ transformations:
process.parent.executable: process.parent.executable.caseless
process.command_line: process.command_line.caseless
process.parent.command_line: process.parent.command_line.caseless
file.path: file.path.caseless
# file_activity: playbook-only pseudo-category spanning all file operations.
- id: playbook_file_activity_add-fields
type: add_condition
conditions:
event.category: 'file'
rule_conditions:
- type: logsource
category: file_activity
# EventType -> event.action is safe as a projection only — the values differ
# (Sysmon DeleteFile/SetValue vs ECS deletion/modification).
- id: playbook_file_activity_event_type
type: field_name_mapping
mapping:
winlog.event_data.EventType: event.action
rule_conditions:
- type: logsource
category: file_activity
- id: playbook_file_event_event_type
type: field_name_mapping
mapping:
winlog.event_data.EventType: event.action
rule_conditions:
- type: logsource
category: file_event
- id: playbook_file_delete_event_type
type: field_name_mapping
mapping:
winlog.event_data.EventType: event.action
rule_conditions:
- type: logsource
category: file_delete
- id: playbook_file_rename_event_type
type: field_name_mapping
mapping:
winlog.event_data.EventType: event.action
rule_conditions:
- type: logsource
category: file_rename
- id: playbook_registry_set_event_type
type: field_name_mapping
mapping:
winlog.event_data.EventType: event.action
rule_conditions:
- type: logsource
category: registry_set
# WMI-Activity events carry no event.category; event.dataset is source-specific
- id: playbook_wmi_activity_add-fields
type: add_condition
conditions:
event.dataset: 'wmi.operational'
rule_conditions:
- type: logsource
category: wmi
- id: playbook_wmi_event_activity_add-fields
type: add_condition
conditions:
event.dataset: 'wmi.operational'
rule_conditions:
- type: logsource
category: wmi_event
- id: playbook_ps_script_field-mapping
type: field_name_mapping
mapping:
ScriptBlockText: powershell.file.script_block_text
Path: file.path
rule_conditions:
- type: logsource
category: ps_script
- id: playbook_ps_script_add-fields
type: add_condition
conditions:
event.dataset: 'windows.powershell_operational'
event.code: '4104'
rule_conditions:
- type: logsource
category: ps_script
# Data (raw EID 400/600 message blob) -> process.command_line: the winlog
# integration parses HostApplication into it; `message` is analyzed text and
# wildcard matches on it silently return zero. No .caseless on this dataset.
- id: playbook_ps_classic_start_field-mapping
type: field_name_mapping
mapping:
Data: process.command_line
rule_conditions:
- type: logsource
category: ps_classic_start
- id: playbook_ps_classic_start_add-fields
type: add_condition
conditions:
event.dataset: 'windows.powershell'
event.code: '400'
rule_conditions:
- type: logsource
category: ps_classic_start
- id: playbook_ps_classic_provider_start_field-mapping
type: field_name_mapping
mapping:
Data: process.command_line
rule_conditions:
- type: logsource
category: ps_classic_provider_start
- id: playbook_ps_classic_provider_start_add-fields
type: add_condition
conditions:
event.dataset: 'windows.powershell'
event.code: '600'
rule_conditions:
- type: logsource
category: ps_classic_provider_start
# Alert docs nest the host under event_data.host.name; no top-level host.name.
- id: playbook_alert_host_field
type: field_name_mapping
mapping:
host.name: event_data.host.name
rule_conditions:
- type: logsource
category: alert
+206 -3
View File
@@ -131,13 +131,22 @@ transformations:
type: field_name_mapping
mapping:
winlog.event_data.Details: registry.data.strings
# field rename only; EventType values (SetValue/CreateKey) still differ from
# event.action values (modification/creation)
winlog.event_data.EventType: event.action
rule_conditions:
- type: logsource
product: windows
category: registry_set
# ecs_windows renames Initiated -> network.direction without translating the value,
# so rules compile to network.direction:"true" and never match (field holds
# ingress/egress). network.initiated carries the boolean on both Defend and Sysmon.
# Keyed on network.direction because ecs_windows has already renamed by this layer.
- id: ecs_fix_network_connection_initiated
type: field_name_mapping
mapping:
network.direction: network.initiated
rule_conditions:
- type: logsource
product: windows
category: network_connection
- id: ecs_fix_image_load
type: field_name_mapping
mapping:
@@ -150,6 +159,30 @@ transformations:
- type: logsource
product: windows
category: image_load
# Defend reports the driver image in dll.path (file.path is null on driver events) —
# same renames as image_load.
- id: ecs_fix_driver_load
type: field_name_mapping
mapping:
file.path: dll.path
file.code_signature.signed: dll.code_signature.exists
winlog.event_data.Signature: dll.code_signature.subject_name
file.code_signature.status: dll.code_signature.status
winlog.event_data.Hashes: dll.hash.sha256
rule_conditions:
- type: logsource
product: windows
category: driver_load
- id: ecs_fix_file_rename
type: field_name_mapping
mapping:
# Defend records the pre-rename path in file.Ext.original.path; Sysmon's
# SourceFilename has no ECS equivalent.
winlog.event_data.SourceFilename: file.Ext.original.path
rule_conditions:
- type: logsource
product: windows
category: file_rename
- id: linux_security_add-fields
type: add_condition
conditions:
@@ -323,6 +356,33 @@ transformations:
rule_conditions:
- type: logsource
category: file_event
# Without event.type scoping, file_delete rules also match creations.
- id: endpoint_file_delete_add-fields
type: add_condition
conditions:
event.category: 'file'
event.type: 'deletion'
rule_conditions:
- type: logsource
category: file_delete
# Defend reports renames as event.action:rename with event.type:change.
- id: endpoint_file_rename_add-fields
type: add_condition
conditions:
event.category: 'file'
event.action: 'rename'
rule_conditions:
- type: logsource
category: file_rename
# event.type:change selects writes and excludes Defend's read-only registry events.
- id: endpoint_registry_set_add-fields
type: add_condition
conditions:
event.category: 'registry'
event.type: 'change'
rule_conditions:
- type: logsource
category: registry_set
# Scope image_load rules to Elastic Endpoint library events (event.category:library, dll.*
# populated).
- id: endpoint_image_load_add-fields
@@ -351,6 +411,149 @@ transformations:
rule_conditions:
- type: logsource
category: network_connection
# Scope on lookup actions, not network.protocol:dns (also matches Zeek port-53).
- id: endpoint_dns_query_add-fields
type: add_condition
conditions:
event.category: 'network'
event.action:
- 'lookup_requested'
- 'lookup_result'
rule_conditions:
- type: logsource
category: dns_query
# OS gates: without them a `product: windows` rule in these categories also matches
# Linux/macOS. Separate conditions (not `product:` on the mappings above) so the
# mappings keep scoping product-less rules.
- id: endpoint_file_create_windows_os-gate
type: add_condition
conditions:
host.os.type: 'windows'
rule_conditions:
- type: logsource
category: file_event
product: windows
- id: endpoint_file_create_linux_os-gate
type: add_condition
conditions:
host.os.type: 'linux'
rule_conditions:
- type: logsource
category: file_event
product: linux
- id: endpoint_file_create_macos_os-gate
type: add_condition
conditions:
host.os.type: 'macos'
rule_conditions:
- type: logsource
category: file_event
product: macos
- id: endpoint_file_delete_windows_os-gate
type: add_condition
conditions:
host.os.type: 'windows'
rule_conditions:
- type: logsource
category: file_delete
product: windows
- id: endpoint_file_delete_linux_os-gate
type: add_condition
conditions:
host.os.type: 'linux'
rule_conditions:
- type: logsource
category: file_delete
product: linux
- id: endpoint_file_delete_macos_os-gate
type: add_condition
conditions:
host.os.type: 'macos'
rule_conditions:
- type: logsource
category: file_delete
product: macos
- id: endpoint_file_rename_windows_os-gate
type: add_condition
conditions:
host.os.type: 'windows'
rule_conditions:
- type: logsource
category: file_rename
product: windows
- id: endpoint_file_rename_linux_os-gate
type: add_condition
conditions:
host.os.type: 'linux'
rule_conditions:
- type: logsource
category: file_rename
product: linux
- id: endpoint_file_rename_macos_os-gate
type: add_condition
conditions:
host.os.type: 'macos'
rule_conditions:
- type: logsource
category: file_rename
product: macos
- id: endpoint_network_connection_windows_os-gate
type: add_condition
conditions:
host.os.type: 'windows'
rule_conditions:
- type: logsource
category: network_connection
product: windows
- id: endpoint_network_connection_linux_os-gate
type: add_condition
conditions:
host.os.type: 'linux'
rule_conditions:
- type: logsource
category: network_connection
product: linux
- id: endpoint_network_connection_macos_os-gate
type: add_condition
conditions:
host.os.type: 'macos'
rule_conditions:
- type: logsource
category: network_connection
product: macos
- id: endpoint_dns_query_windows_os-gate
type: add_condition
conditions:
host.os.type: 'windows'
rule_conditions:
- type: logsource
category: dns_query
product: windows
- id: endpoint_dns_query_linux_os-gate
type: add_condition
conditions:
host.os.type: 'linux'
rule_conditions:
- type: logsource
category: dns_query
product: linux
- id: endpoint_dns_query_macos_os-gate
type: add_condition
conditions:
host.os.type: 'macos'
rule_conditions:
- type: logsource
category: dns_query
product: macos
# registry_set / image_load / driver_load are not gated — windows-only telemetry.
# wmi / wmi_event are not scoped here: their only handle is event.dataset
- id: endpoint_driver_load_add-fields
type: add_condition
conditions:
event.category: 'driver'
rule_conditions:
- type: logsource
category: driver_load
# Maps "alert" category to SO Alert events
- id: alert_so_add-fields
type: add_condition
+1 -1
View File
@@ -88,7 +88,7 @@
'host': GLOBALS.manager_ip,
'password': PG_PASS,
'port': 5432,
'sslMode': 'allow',
'sslMode': 'require',
'user': PG_USER,
}
}) %}
+6 -3
View File
@@ -486,7 +486,7 @@ soc:
global: True
advanced: True
sslMode:
description: "Use encrypted connections to the PostgreSQL server. Must be one of the following values: disable, allow, prefer, require, verify-ca, verify-full. Defaults to allow."
description: "Use encrypted connections to the PostgreSQL server. Must be one of the following values: disable, allow, prefer, require, verify-ca, verify-full."
global: True
advanced: True
database:
@@ -783,8 +783,11 @@ soc:
Orchestrator:
description: The initial agent in most agentic conversations. This agent will delegate requests to specialized agents.
global: True
Hunter:
description: This agent is specialized in querying events.
Investigator:
description: This agent investigates alerts, explains events and records, and hunts through event data. It can also acknowledge alerts and escalate to cases.
global: True
Detection Engineer:
description: This agent manages detections and their overrides, including tuning noisy rules and authoring rule content.
global: True
client:
assistant: