Cyber Threat Brief — August 24 2026

⚠️ This report is AI-generated. Always validate findings.

1. Zimbra SNMP Command Injection — CVE-2026-73570

TL;DR: Unauthenticated RCE in Zimbra Collaboration via SNMP notification processing, exploited ITW per CERT Polska. CISA KEV federal deadline is TODAY (August 24).

What’s New:

  • CISA added to KEV August 21; CERT Polska confirmed active exploitation August 17
  • CWE-78 OS command injection in SNMP notification handler — attacker sends crafted SMTP requests, gets code execution as zimbra user
  • Affects ZCS versions before 10.1.20 when zimbra-snmp package is installed and SNMP notifications enabled
  • CVSS 8.9; no authentication required, no user interaction
  • Patched in ZCS 10.1.20 (released July 20 2026)

Actionable Intel

ArtifactTypeATT&CKLog SourceAction
Crafted SMTP payloads triggering SNMP notification pathInitial AccessT1190Zimbra MTA logs, /var/log/zimbra.logHunt for anomalous service-status messages (unknown payloads changing state)
Unexpected processes spawned as zimbra userExecutionT1059.004Sysmon (Linux), auditdAlert on shell/child processes from Zimbra service accounts
Web shells in Zimbra webrootPersistenceT1505.003File integrity monitoringSweep /opt/zimbra/jetty/webapps/ and /opt/zimbra/httpd/ for new .jsp/.php files
Cron jobs or systemd units created by zimbra userPersistenceT1053.003auditd, cron logsHunt for new scheduled tasks under zimbra context
Outbound C2 from Zimbra serverC2T1071.001Firewall/proxy logsBaseline and alert on outbound connections from mail infrastructure

Detection

SourceRuleGap
Splunk ESCUNoneNo coverage for Zimbra SNMP injection; need custom rule on zimbra.log anomalous service-status patterns
ElasticNoneNo Zimbra-specific rules; gap for SMTP-triggered command injection
SigmaGeneric web shell detection rules applyMissing Zimbra-specific SNMP notification abuse detection

Sources: CISA KEV · The Hacker News · Security Arsenal · SOCRadar


2. GitLab GraphQL Code Injection — CVE-2026-19478

TL;DR: Critical code injection (CVSS 9.4) in self-managed GitLab via GraphQL directive lets unauthenticated attackers modify/delete public projects. Exploited ITW within 2 days of disclosure.

What’s New:

  • Disclosed August 17, out-of-band patches released same day; WatchTowr confirmed exploitation by August 20
  • Unauthenticated attacker abuses GraphQL directive to inject code — can delete repositories, forge merge records, ban maintainers
  • Affects GitLab CE/EE across multiple 18.x and 19.x branches
  • Horizon3.ai published detailed attack research with exploitation methodology
  • Fixed in 19.2.4, 19.1.6, 19.0.8, 18.11.11

Actionable Intel

ArtifactTypeATT&CKLog SourceAction
GraphQL requests containing @gl_introduced directiveExploitationT1190GitLab web/API logs, WAFHunt/block requests with @gl_introduced in GraphQL queries
Unauthorized project deletions or merge record modificationsImpactT1485GitLab audit logs (audit_events table)Alert on project deletions or merges by unauthenticated sessions
Mass repository enumeration via GraphQLReconT1526GitLab API rate-limit logsAlert on high-volume GraphQL project queries from single source
Maintainer account bans from non-admin sessionsPrivilege abuseT1531GitLab admin audit logAlert on account status changes outside admin context

Detection

SourceRuleGap
Splunk ESCUNoneNo GitLab-specific coverage; need custom search on GitLab audit/API logs for @gl_introduced
ElasticNoneNo GitLab GraphQL injection rules
SigmaNoneNo coverage; custom rule needed for GraphQL directive injection patterns in web logs

Sources: SecurityWeek · The Hacker News · Horizon3.ai · Security Affairs


3. Metabase Password-Reset SQL Injection — CVE-2026-72898

TL;DR: CVSS 10.0 unauthenticated SQLi in Metabase /reset_password endpoint grants admin access to all connected data warehouses (Snowflake, BigQuery, Redshift, etc.). Five companies breached before disclosure; CISA KEV since August 11.

What’s New:

  • Unauthenticated SQL injection via /reset_password endpoint — attacker gains admin access to Metabase instance
  • Admin access exposes connection strings and credentials for every connected database/warehouse
  • Five companies confirmed breached before Metabase disclosed August 6; KEV added August 11
  • Downstream risk is extreme: a single Metabase compromise pivots to Databricks, MongoDB, Oracle, Snowflake, Redshift, BigQuery
  • Federal remediation deadline was August 14 (passed)

Actionable Intel

ArtifactTypeATT&CKLog SourceAction
Anomalous requests to /api/session/reset_passwordInitial AccessT1190Metabase application logs, WAF/reverse proxy logsHunt for SQLi payloads in password reset requests
New admin accounts created without corresponding user registrationPersistenceT1136.001Metabase audit log, database core_user tableAlert on admin account creation outside normal workflows
Bulk database credential access from Metabase admin panelCredential AccessT1552.001Metabase application logsAlert on mass connection-string reads
Unexpected queries to connected warehouses from Metabase service accountLateral MovementT1021Snowflake/BigQuery/Redshift audit logsBaseline Metabase query patterns; alert on anomalies

Detection

SourceRuleGap
Splunk ESCUNoneNo Metabase-specific coverage; need custom rule on WAF/proxy logs for SQLi in /reset_password
ElasticNoneGeneric SQLi detection may partially catch; no Metabase-specific rule
SigmaGeneric SQL injection web rulesMissing Metabase endpoint-specific detection

Sources: The Hacker News · GBHackers · TechTimes · runZero


Status Updates

  • CVE-2026-72529/72530 (TrueConf Server): Kaspersky published full IOC dump — 20+ MD5 hashes (PhantomCore, PhantomGraph DLLs, web shells, ELF backdoors, rootkit), 5 C2 IPs, 10 domains, YARA rules, and Windows service names (SysExcSvc, SysReadSvc). Federal deadline passed August 23. Original brief. Kaspersky Securelist.
  • CVE-2026-33824 (Windows IKE): Public PoC exploit code now on GitHub (kaleth4/CVE-2026-33824); claims reliable execution across patch table. Prioritize patching any remaining unpatched IKE/VPN endpoints. Original brief.
  • CVE-2026-69836 (Microsoft Entra ID): Microsoft corrected advisory — vulnerability was not exploited ITW; mistakenly flagged. Fully mitigated server-side, no customer action needed. Downgrade priority. Original brief.