Rapid7's latest Metasploit Wrap Up confirms what defenders running Zimbra Collaboration Suite feared: the identity-boundary vulnerabilities documented in Rapid7's Breaking Zimbra Identity Boundaries, Part One research are now weaponized in the world's most common exploitation framework. Community contributor jburgess-r7 landed three pull requests: a new shared mixin, Msf::HTTP::Zimbra, that standardizes account/admin SOAP authentication, canonical identity resolution, mailbox-server lookup, and version parsing for every Zimbra module going forward — and, built on top of it, working exploit modules for the CVEs from that research track, including CVE-2026-73571, an authorization bypass in Zimbra's SaveDraftRequest SOAP handling that lets an ordinary authenticated user send mail as another identity.
For defenders, the operational picture is simple: the barrier to exploiting this flaw just dropped to zero. Any attacker with a single low-privilege Zimbra account — phished credentials, a compromised contractor mailbox, a brute-forced webmail login — can now impersonate arbitrary users on the platform. That means CEO fraud from your own infrastructure, internal phishing that bypasses every external-sender warning, and attribution chaos during incident response. If you run Zimbra, assume exploit attempts will follow within days, not weeks.
Technical Analysis
Affected Component
The vulnerability lives in Zimbra Collaboration Suite's SOAP API, specifically in the handling of SaveDraftRequest. Zimbra's SOAP endpoint (/service/soap on the mailbox/proxy nodes) accepts draft-save operations that can carry sender identity attributes. The flaw is a broken authorization check: the server fails to validate that the authenticated session is actually entitled to use the sender identity specified in the request. An attacker supplies a victim's email address (or an identity they do not own) in the draft, and Zimbra accepts, stores, and ultimately sends the message under that identity.
Exploitation Requirements
- One valid, low-privilege Zimbra account. No admin rights, no special roles. Any mailbox user suffices.
- Network reachability to the Zimbra web/SOAP interface (typically TCP 443 or 7071).
- That's it. The Metasploit module handles the SOAP plumbing — authentication, mailbox-server lookup, and canonical identity resolution — automatically via the new shared mixin.
Why the Shared Mixin Matters Defensively
The Msf::HTTP::Zimbra mixin is arguably the more significant long-term development. By abstracting SOAP authentication, identity resolution, mailbox-server discovery, and version parsing into a common foundation, Rapid7 has made it trivial to build future Zimbra modules quickly. Expect additional modules covering the rest of the identity-boundary research track to land in subsequent releases. This is a signal to treat Zimbra as an actively researched attack surface, not a one-CVE event.
Impact Chain from a Defender's View
- Attacker authenticates with any valid Zimbra credential.
- Attacker issues a crafted
SaveDraftRequestnaming a victim identity (CEO, CFO, HR, a privileged admin). - Zimbra stores and sends the message with the victim's address as the sender — no spoofing headers, no external gateway flags, no DMARC failure. The mail originates from your legitimate infrastructure.
- Result: internal business email compromise (BEC), credential harvesting campaigns that pass all external-sender banners, and poisoned forensic attribution (your logs will show the victim's identity on outbound mail).
Exploitation Status
- Public exploit code: Yes — a working Metasploit module is merged into the framework.
- Complexity: Low. Authenticated, single-request class exploitation with framework support.
- CISA KEV: Not listed at time of writing, but the combination of public exploit, low privilege requirements, and Zimbra's history of being targeted by APT groups makes rapid weaponization a near-certainty. Zimbra flaws have historically been added to KEV quickly once public PoCs exist.
Detection & Response
The reliable detection surface here is Zimbra's own logging, not the endpoint. mailbox.log records SOAP operations and the authenticated account context; nginx access logs (/opt/zimbra/log/nginx.access.log) capture the request path, source IP, and user agent; and Zimbra's audit logging ties sessions to accounts. The highest-fidelity signal is a sender identity that does not match the authenticated account on draft save / send operations, and session behavior consistent with the Metasploit mixin (rapid SOAP authentication followed immediately by SaveDraft/SendMsg operations, often from a non-interactive client).
---
title: Zimbra SOAP SaveDraftRequest Identity Mismatch (CVE-2026-73571)
id: 3f8a2c61-7b4d-4e9a-b1c5-9d2e6f0a8347
status: experimental
description: Detects potential exploitation of CVE-2026-73571 where a Zimbra SOAP SaveDraft or SendMsg operation is issued with a sender identity that differs from the authenticated account, indicating authorization bypass abuse.
references:
- https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-breaking-zimbra-s-identity-boundaries
author: Security Arsenal
date: 2026/05/08
tags:
- attack.exploitation_public_facing_application
- attack.t1190
logsource:
product: linux
service: zimbra
detection:
selection_op:
soap_op|contains:
- 'SaveDraftRequest'
- 'SendMsgRequest'
filter_matching_identity:
identity_match: 'true'
condition: selection_op and not filter_matching_identity
falsepositives:
- Legitimate delegated mailbox access (shared inboxes, 'send on behalf of' configurations) - baseline known delegate pairs and exclude
- Zimbra mobile/activesync clients using persona features - inventory known personas
level: high
---
title: Non-Interactive Zimbra SOAP Authentication Burst Followed by Draft Operations
id: 8c1d4e72-3a5f-4b68-9d21-6e7f0c2b9518
status: experimental
description: Detects SOAP session behavior consistent with automated tooling (including the Metasploit Zimbra mixin) - rapid AuthRequest immediately followed by SaveDraft/SendMsg operations from a single source without webmail UI request patterns.
references:
- https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-breaking-zimbra-s-identity-boundaries
author: Security Arsenal
date: 2026/05/08
tags:
- attack.exploitation_public_facing_application
- attack.t1190
- attack.t1078
logsource:
category: webserver
detection:
selection:
cs-uri-stem|contains: '/service/soap'
cs-method: 'POST'
filter_ui:
cs-user-agent|contains:
- 'Mozilla'
- 'AppleWebKit'
- 'ZimbraMobileSync'
- 'Outlook'
condition: selection and not filter_ui
falsepositives:
- Legitimate SOAP API integrations (archiving, CRM connectors, migration tools) - whitelist known service account user agents and source IPs
level: medium
---
title: Mass Outbound Mail From Single Zimbra Account Using Multiple Sender Identities
id: 5b9e2f04-1d3a-4c76-8a45-2f6b9d0e1732
status: experimental
description: Detects a single authenticated Zimbra session or source IP generating outbound messages across multiple distinct From identities, a strong indicator of identity-boundary abuse for internal phishing or BEC staging.
references:
- https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-breaking-zimbra-s-identity-boundaries
author: Security Arsenal
date: 2026/05/08
tags:
- attack.t1078
- attack.t1566
logsource:
product: linux
service: zimbra
detection:
selection:
event_type: 'outbound_mail'
condition: selection | count(distinct from_identity) by source_ip, session_id > 3
falsepositives:
- Service accounts legitimately sending on behalf of multiple system identities (notification systems) - maintain an allowlist
level: high
For Sentinel/Defender environments ingesting Zimbra logs via Syslog/CEF forwarders, the following hunt query pivots on the same identity-mismatch logic and adds a burst-detection lens for automated SOAP tooling:
// Hunt: Zimbra SOAP SaveDraft/SendMsg with sender identity mismatch (CVE-2026-73571)
// Assumes Zimbra mailbox.log and nginx access logs are forwarded via Syslog/CEF
let Lookback = 7d;
let SoapOps = dynamic(["SaveDraftRequest", "SendMsgRequest"]);
Syslog
| where TimeGenerated > ago(Lookback)
| where Computer has_any ("zimbra", "mailbox", "zcs") or Facility == "mail"
| where SyslogMessage has "soap" and SyslogMessage has_any (SoapOps)
| parse SyslogMessage with * "name=" AuthenticatedAccount ";" *
| parse SyslogMessage with * "from=" SenderIdentity ";" *
| where isnotempty(AuthenticatedAccount) and isnotempty(SenderIdentity)
| where SenderIdentity !contains AuthenticatedAccount
| extend Suspicious = true
| summarize Operations = count(),
DistinctSenders = dcount(SenderIdentity),
SenderList = make_set(SenderIdentity),
FirstSeen = min(TimeGenerated),
LastSeen = max(TimeGenerated)
by AuthenticatedAccount, Computer, bin(TimeGenerated, 1h)
| where DistinctSenders > 1 or Operations > 20
| sort by LastSeen desc
// Hunt: Automated/non-browser clients hitting Zimbra SOAP endpoint (Metasploit mixin behavior)
// Via CommonSecurityLog if Zimbra sits behind a proxy/WAF forwarding to Sentinel
CommonSecurityLog
| where TimeGenerated > ago(3d)
| where RequestURL contains "/service/soap"
| where RequestMethod == "POST"
| where RequestUserAgent !has_any ("Mozilla", "AppleWebKit", "Zimbra", "Outlook", "curl/7.8")
or isempty(RequestUserAgent)
| summarize Requests = count(),
DistinctURIs = dcount(RequestURL),
UserAgents = make_set(RequestUserAgent)
by SourceIP, DestinationHostName, bin(TimeGenerated, 15m)
| where Requests > 10
| sort by Requests desc
Velociraptor is warranted if you suspect a Zimbra mailbox node itself was touched (e.g., the identity bypass was chained with host-level access, or you need to verify log integrity before trusting it). This artifact pulls the relevant log evidence directly off the server:
-- Artifact: Linux.Hunt.ZimbraIdentityAbuse
-- Collects Zimbra SOAP/mailbox log evidence of SaveDraft identity mismatch
-- and outbound mail anomalies for CVE-2026-73571 triage.
LET logs <= SELECT FullPath, Size, Mtime
FROM glob(globs=[
'/opt/zimbra/log/mailbox.log*',
'/opt/zimbra/log/nginx.access.log*',
'/opt/zimbra/log/audit.log*',
'/opt/zimbra/log/zmmailboxd.out'
])
ORDER BY Mtime DESC
LIMIT 50
LET soap_hits <= SELECT FullPath, Line
FROM foreach(
row={
SELECT FullPath FROM logs WHERE FullPath =~ 'mailbox'
},
query={
SELECT FullPath, Line
FROM parse_lines(filename=FullPath)
WHERE Line =~ 'SaveDraftRequest|SendMsgRequest'
AND Line =~ 'oip='
})
SELECT FullPath,
Size,
Mtime,
get(item=soap_hits, field='Line') AS SoapEvidence
FROM logs
Triage Guidance
- Validate against delegated access first. Shared mailboxes and "send on behalf of" configurations are the primary false-positive source. Export your delegate/grant mappings (
zmprov) before you start paging anyone. - Correlate with nginx source IPs. A mismatch event from an IP with no prior webmail history for that account is far more suspicious than one from the user's normal egress.
- Check outbound queues. Review
/opt/zimbra/log/zimbra.log(postfix) for messages sent under victim identities — thefrom=in postfix logs reflects what was actually transmitted.
Remediation
1. Patch Immediately
Apply the current Zimbra security patch for your release train. Zimbra publishes fixes via its security advisory portal and per-release patch bundles (10.x, 9.x, and 8.8.15 lines). Pull the exact patch version for your deployment from the official advisories:
- Zimbra Security Center: https://www.zimbra.com/product/security/
- Zimbra Security Advisories: https://wiki.zimbra.com/wiki/Security_Center
- Rapid7 research (technical detail): https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-breaking-zimbra-s-identity-boundaries
Do not rely on a Web Application Firewall rule as a substitute for the patch — the exploit uses legitimate SOAP structure, and the difference between valid and malicious traffic is semantic (the identity claim), not syntactic.
2. Verify and Hunt with This Script
#!/bin/bash
# Zimbra CVE-2026-73571 verification and hunting script
# Run on each Zimbra mailbox node. Requires zimbra or root context.
echo "=== [1] Zimbra version check ==="
su - zimbra -c 'zmcontrol -v'
echo ""
echo "=== [2] Installed patches ==="
su - zimbra -c 'zmprov -l gs $(zmhostname) zimbraPatchVersion 2>/dev/null' || echo "Check patch level via admin console"
echo ""
echo "=== [3] SaveDraft/SendMsg SOAP operations in last 7 days ==="
grep -h -E 'SaveDraftRequest|SendMsgRequest' /opt/zimbra/log/mailbox.log* 2>/dev/null | \
grep -oE 'oip=[0-9.]+|name=[^;,]+' | sort | uniq -c | sort -rn | head -30
echo ""
echo "=== [4] Non-browser user agents hitting /service/soap (nginx) ==="
grep -h '/service/soap' /opt/zimbra/log/nginx.access.log* 2>/dev/null | \
grep -vE 'Mozilla|AppleWebKit|ZimbraMobile|Outlook' | \
awk '{print $1}' | sort | uniq -c | sort -rn | head -20
echo ""
echo "=== [5] Outbound mail From: distribution (last 24h) - look for identity sprawl ==="
grep -h 'from=<' /var/log/zimbra.log 2>/dev/null | \
grep -oE 'from=<[^>]+>' | sort | uniq -c | sort -rn | head -30
echo ""
echo "=== [6] Delegated grant inventory (baseline for FP tuning) ==="
su - zimbra -c 'for acct in $(zmprov -l gaa); do zmprov gg $acct 2>/dev/null; done' | head -50
3. Compensating Controls While You Patch
- Restrict SOAP endpoint exposure. If your architecture allows it, restrict
/service/soapat the proxy/WAF to known integration source IPs, forcing interactive users through the web client path. Test carefully — ActiveSync and some mobile flows use SOAP. - Enforce MFA on webmail and review active sessions; expire all auth tokens (
zmprov fc -a allstyle cache flushes / session invalidation) after patching to kill any attacker-held sessions. - Tighten identity controls. Audit persona and delegate configurations — the exploit's value scales with how many high-value identities exist on the platform.
- Outbound alerting. Stand up detection on any single account generating mail under more than N distinct From identities per hour (the third Sigma rule above).
4. Incident Response Triggers
If your hunt returns hits: preserve mailbox.log, audit.log, nginx.access.log, and zimbra.log before rotation; enumerate every message sent under mismatched identities and pull recipients; treat any internal recipient engagement (replies, credential submissions) as a BEC compromise chain; and rotate credentials for both the attacking account and every impersonated identity.
The Bigger Picture
The Zimbra identity-boundary research track — and Rapid7's decision to build shared infrastructure around it — tells you where attacker and researcher attention is heading: the authorization logic inside collaboration platforms, not just their memory-corruption bugs. Identity-boundary flaws are quiet, leave minimal artifacts, and produce mail that every one of your controls trusts. Patch CVE-2026-73571, but also use this moment to build the identity-mismatch detection muscle described above, because the next modules built on Msf::HTTP::Zimbra are already coming.
Related Resources
Security Arsenal Penetration Testing Services AlertMonitor Platform Book a SOC Assessment vulnerability-management Intel Hub
Is your security operations ready?
Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.