Email Infrastructure Compromise and Impersonation

High severity TLP:CLEAR Research by Rapid7 Published 21 Nov 2026

Hypothesis

An adversary has exploited vulnerabilities in Zimbra, Cisco Secure Email Gateway, or BeyondTrust to achieve remote code execution and persistent access through script overwrites and identity boundary bypasses.

Brief

Why Now

The Metasploit Wrap Up: Breaking Zimbra's Identity Boundaries (https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-breaking-zimbra-s-identity-boundaries) describes how attackers exploit vulnerabilities in mail appliances to bypass security controls. These platforms sit on the network edge and often lack endpoint detection agents. When an adversary compromises a gateway, they gain the ability to impersonate users and read sensitive data without triggering traditional identity alerts.

How the Hunt Flows

The hunt starts with a scoping query. It identifies appliance hosts running versions vulnerable to the targeted CVEs, including recent flaws in Zimbra and Cisco SEG. This phase ensures the analyst focuses only on systems with a known exposure. The next phase involves parallel queries to identify early access indicators. The first query searches for authentication anomalies on Zimbra GraphQL and SOAP endpoints. It flags high-frequency sign-in attempts from external IP addresses, which indicates a rate-limit bypass. The second query examines HTTP activity for specific request paths. It looks for exploit-related handlers like SaveDraftRequest or proprietary appliance APIs used during payload delivery. An agent then reviews these results to provide a suspicious or malicious verdict per host. This prevents the analyst from chasing every failed scan and highlights systems where exploit traffic likely succeeded. The hunt then pivots to persistence. It checks for file activity involving the core_watch.sh script, which attackers often overwrite on Cisco SEG appliances. Simultaneously, it searches for process activity where the system cron daemon executes this script. Finding this script running with root privileges confirms that an adversary establishes persistence.

What the Hunt Cannot See

This hunt has two main limitations. It requires detailed HTTP access logs exported to the central logging platform. If the appliance does not log specific URL paths or SOAP headers, the hunt cannot distinguish between malicious impersonation and legitimate traffic. Additionally, if an attacker executes a malicious script and quickly reverts the file, the hunt might miss the file modification event unless the appliance has real-time endpoint auditing enabled.

Steps

  1. Query, scoping

    Identify vulnerable mail infrastructure

    Find appliance hosts running versions vulnerable to the targeted CVEs to narrow the focus of the hunt.

    sql, reads hb_vulnerability_finding
    SELECT resource_uid AS target, cve_uid, severity, affected_package_name, affected_package_versionFROM hb_vulnerability_findingWHERE instr(',' || '{{target_cves}}' || ',', ',' || cve_uid || ',') > 0  AND status != 'suppressed'

    What a hit looks like. A list of resource IDs (hostnames or IPs) currently exposed to the targeted vulnerabilities. Silence means no known vulnerable appliances are currently reporting.

  2. Query, baseline

    Zimbra GraphQL auth anomalies

    Detect rate-limit bypass and credential stuffing on Zimbra Network Edition GraphQL endpoints.

    sql, reads hb_auth_signin
    SELECT device_hostname, src_endpoint_ip, request_uri, COUNT(*) AS attemptsFROM hb_auth_signinWHERE (LOWER(request_uri) LIKE '%/graphql%' OR LOWER(event_type) LIKE '%soap%')  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')GROUP BY device_hostname, src_endpoint_ip, request_uriHAVING attempts > 20

    What a hit looks like. High volumes of auth attempts on GraphQL from external IPs indicate a rate-limit bypass attempt.

  3. Query, detection candidate

    Exploit request paths

    Find web requests targeting known vulnerable handlers for Zimbra, Cisco SEG, and BeyondTrust.

    sql, reads hb_http_activity
    SELECT device_hostname, src_endpoint_ip, url_path, url_query, status_codeFROM hb_http_activityWHERE (LOWER(url_path) LIKE '%savedraftrequest%' OR LOWER(url_path) LIKE '%/mrs/%' OR LOWER(url_path) LIKE '%/ecs/%')  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')

    What a hit looks like. Requests to delegated sending handlers or proprietary appliance APIs from unauthenticated sources.

  4. Agent triage

    Evaluate early access patterns

    Assess whether the observed auth anomalies and HTTP paths indicate successful initial access by an adversary.

  5. Query, triage

    Cisco SEG script overwrites

    Identify the overwriting of core_watch.sh as reported in Cisco SEG CVE-2026-76519 exploitation.

    sql, reads hb_file_activity
    SELECT device_hostname, file_path, process_name, actor_user_name, timeFROM hb_file_activityWHERE LOWER(file_path) LIKE '%/core_watch.sh'  AND activity_id IN (1, 3, 5)  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')

    What a hit looks like. Creation or modification of the core_watch.sh script, especially if originating from a web-server or scanning process.

  6. Query, enrichment

    Cron-driven script execution

    Verify if the overwritten script is being executed with root privileges by the system cron daemon.

    sql, reads hb_process_activity
    SELECT device_hostname, process_path, parent_process_name, process_cmd_line, user_name, timeFROM hb_process_activityWHERE (LOWER(process_path) LIKE '%/core_watch.sh' OR LOWER(process_cmd_line) LIKE '%core_watch.sh%')  AND (LOWER(parent_process_name) LIKE '%cron%' OR LOWER(parent_process_name) LIKE '%crond%')  AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0)  AND time >= datetime('now', '-{{lookback_days}} days')

    What a hit looks like. Execution of core_watch.sh by cron verifies successful persistence and elevation to root.

  7. Agent triage

    Evaluate full compromise chain

    Correlate the early access evidence with follow-on persistence to reach a final verdict of compromise.

  8. Decision

    Route remediation

    Determine whether to trigger containment based on the agent's verdict.

  9. Response action

    Isolate compromised host

    Immediately prevent lateral movement and further data exfiltration from the compromised mail appliance.

  10. Analyst task

    Forensic mailbox audit

    Conduct manual investigation to determine the impact of email impersonation or data access.

  11. Analyst task

    Patch and harden infrastructure

    Restore the appliance to a known good, patched state and apply hardening measures.

Coverage

4 of 4 stages covered

Stage and ATT&CK techniqueCoveredHow, or why not
Exploitation of Public-Facing Appliance
T1190
Yes scoping-vulnerable-appliances, zimbra-auth-anomalies, appliance-exploit-request-paths
Persistence via Script Overwrite
T1053.003
Yes cisco-seg-script-overwrite
Execution via Cron Job
T1053.003
Yes cron-driven-script-execution
Identity Boundary Bypass
T1190
Yes appliance-exploit-request-paths, evaluate-early-access

Blind spots

  • Needs detailed HTTP access logs from the appliance exported to hb_http_activity. Proprietary appliances may not export full URL paths or headers by default, making it difficult to distinguish exploit traffic from legitimate SOAP requests. It would answer: whether the specific SaveDraftRequest or MRS payload was delivered Remediation: Enable and ingest raw appliance access logs into the centralized logging platform.
  • Needs real-time process auditing on the appliance endpoint. If auditing is not enabled on the appliance itself, the transient execution of a malicious script may leave no trace in the centralized logs. It would answer: what code the core_watch.sh script executed before it was potentially reverted Remediation: Deploy osquery or a similar auditing agent to all network appliances.

Parameters and data

Parameters

ParameterTypeDefaultWhat it is
lookback_daysnumber14Days of history to examine.
scope_hostslist[host]noneHostnames to restrict queries; leave empty to scan the full estate.
target_cveslist[string]CVE-2026-73571, CVE-2026-93637, CVE-2026-76519, CVE-2026-76461, CVE-2026-102764, CVE-2024-25420CVE IDs associated with appliance exploits.

Data sources

SourceCategoryTelemetry
Endpoint telemetry (hb_ surfaces)endpointendpoint
Identity / sign-in telemetryidentityidentity
Web server / proxy logssiemnetwork

Source

An open hunt.md file. It runs anywhere that reads the format.

hunt.md
---
analysis: A single rule could detect one of these CVEs, but this hunt correlates vulnerability
  status with authentication anomalies, filesystem persistence, and root-level cron
  execution across multiple surfaces, providing the context necessary to confirm a
  full compromise.
blind_spots:
- id: missing-appliance-access-logs
  owner: Infrastructure Team
  question: whether the specific SaveDraftRequest or MRS payload was delivered
  remediation: Enable and ingest raw appliance access logs into the centralized logging
    platform.
  requires: detailed HTTP access logs from the appliance exported to hb_http_activity
  risk: Proprietary appliances may not export full URL paths or headers by default,
    making it difficult to distinguish exploit traffic from legitimate SOAP requests.
  stage: initial-access-appliance-exploit
- id: ephemeral-script-execution
  owner: Security Engineering
  question: what code the core_watch.sh script executed before it was potentially
    reverted
  remediation: Deploy osquery or a similar auditing agent to all network appliances.
  requires: real-time process auditing on the appliance endpoint
  risk: If auditing is not enabled on the appliance itself, the transient execution
    of a malicious script may leave no trace in the centralized logs.
  stage: execution-via-cron-job
coverage:
- stage: initial-access-appliance-exploit
  status: covered
  steps:
  - scoping-vulnerable-appliances
  - zimbra-auth-anomalies
  - appliance-exploit-request-paths
- stage: persistence-via-script-overwrite
  status: covered
  steps:
  - cisco-seg-script-overwrite
- stage: execution-via-cron-job
  status: covered
  steps:
  - cron-driven-script-execution
- stage: identity-boundary-bypass
  status: covered
  steps:
  - appliance-exploit-request-paths
  - evaluate-early-access
guardrails:
  claims: no_unsupported
  evidence: citation_required
  missing_data: not_benign
  telemetry: untrusted
hunt:
  applicability: campaign-specific
  handoff: keep-as-periodic-hunt
  justification: Compromise of email gateways allows for RCE and silent impersonation,
    bypassing traditional identity controls and enabling large-scale spear-phishing
    or data theft.
  methodology: model-assisted
  trigger: intel-report
hypothesis: An adversary has exploited vulnerabilities in Zimbra, Cisco Secure Email
  Gateway, or BeyondTrust to achieve remote code execution and persistent access through
  script overwrites and identity boundary bypasses.
labels:
- hunt
- attack.t1190
- attack.t1053.003
- attack.t1021.001
- execution
- initial access
- lateral movement
- persistence
name: Email Infrastructure Compromise and Impersonation
parameters:
  lookback_days:
    default: '14'
    description: Days of history to examine.
    type: number
  scope_hosts:
    default: []
    description: Hostnames to restrict queries; leave empty to scan the full estate.
    type: list[host]
  target_cves:
    default:
    - CVE-2026-73571
    - CVE-2026-93637
    - CVE-2026-76519
    - CVE-2026-76461
    - CVE-2026-102764
    - CVE-2024-25420
    description: CVE IDs associated with appliance exploits.
    from:
      kind: article
      observed: '2026-11-20'
      ref: Rapid7 Research
    type: list[string]
provenance:
  authors:
  - name: Huntbase hunt generation
    org: huntbase.io
  generated:
    by: huntbase-hunt-generation
    from: https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-breaking-zimbra-s-identity-boundaries
    gates:
    - dry-run
    - lint
    - critic
    model: hb_google/gemini-3-flash-preview
rationale: Prioritize internet-facing Zimbra servers and Cisco Secure Email Gateways.
  Use the scoping query to populate scope_hosts for behavioural queries.
references:
- name: 'Metasploit Wrap Up: Breaking Zimbra''s Identity Boundaries'
  url: https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-breaking-zimbra-s-identity-boundaries
related:
- hunt: appliance-lateral-movement-rdp
  reason: This hunt focuses on the appliance breach itself; movement into the Windows
    estate via RDP is handled in a separate lateral movement hunt.
  relation: out-of-scope-alternative
scenario:
  stages:
  - name: Exploitation of Public-Facing Appliance
    observables:
    - CVE-2026-76519
    - CVE-2026-76461
    - CVE-2026-102764
    - CVE-2026-93637
    - Zimbra Network Edition GraphQL authentication endpoint
    - BeyondTrust Remote Support MRS
    slug: initial-access-appliance-exploit
    tactic: initial-access
    techniques:
    - T1190
  - name: Persistence via Script Overwrite
    observables:
    - /data/bin/core_watch.sh
    - TAR files containing PAX archives
    - Metasploit payload overwriting system scripts
    slug: persistence-via-script-overwrite
    tactic: persistence
    techniques:
    - T1053.003
  - name: Execution via Cron Job
    observables:
    - core_watch.sh execution every 5 minutes
    - root privilege execution from cron parent process
    slug: execution-via-cron-job
    tactic: execution
    techniques:
    - T1053.003
  - name: Identity Boundary Bypass
    observables:
    - Zimbra SaveDraftRequest SOAP handler
    - CVE-2026-73571
    - admin.authorizedJIDs system property modification
    - impersonation without delegation grant
    slug: identity-boundary-bypass
    tactic: lateral-movement
    techniques:
    - T1190
  summary: Threat actors exploit unauthenticated vulnerabilities in email infrastructure
    like Cisco Secure Email Gateway and Zimbra to achieve persistence and lateral
    movement. The attack chain includes arbitrary file writes to hijack system scripts
    executed by cron jobs and bypassing identity boundaries via SOAP and GraphQL requests.
severity: high
targets:
  analyst:
    name: Tier-2 analyst
    role: analyst
  endpoint:
    category: endpoint
    name: Endpoint telemetry (hb_ surfaces)
    telemetry:
    - endpoint
  hunter:
    agent: true
    name: Hunt agent
  identity:
    category: identity
    name: Identity / sign-in telemetry
    telemetry:
    - identity
  web:
    category: siem
    name: Web server / proxy logs
    telemetry:
    - network
tlp: clear
type: investigation
---


# Email Infrastructure Compromise and Impersonation

This hunt identifies mail appliances exposed to recent high-severity vulnerabilities, then looks for evidence of multi-stage exploitation. It starts by analyzing authentication bypass attempts on Zimbra and exploit request paths for Cisco SEG and BeyondTrust. The second phase corroborates this early access with evidence of persistence through the modification of system scripts like core_watch.sh and their subsequent execution via cron. An agent evaluates the full compromise chain to identify compromised gateways for isolation and forensic review.

## scoping-vulnerable-appliances
<!-- Identify vulnerable mail infrastructure -->
Find appliance hosts running versions vulnerable to the targeted CVEs to narrow the focus of the hunt.

```sqlite target=endpoint role=scoping params=(target_cves=target_cves)
~~~yaml
expected: A list of resource IDs (hostnames or IPs) currently exposed to the targeted
  vulnerabilities. Silence means no known vulnerable appliances are currently reporting.
reads:
- resource_uid
- cve_uid
- severity
- affected_package_name
- affected_package_version
silence: not_evidence_of_absence
source: hb_vulnerability_finding
verified: dry-run
verified_at: '2026-11-21'
~~~
SELECT resource_uid AS target, cve_uid, severity, affected_package_name, affected_package_version FROM hb_vulnerability_finding WHERE instr(',' || '{{target_cves}}' || ',', ',' || cve_uid || ',') > 0 AND status != 'suppressed'
```

## parallel-early-access
<!-- Analyze early stage access indicators -->
parallel:
- → zimbra-auth-anomalies
- → appliance-exploit-request-paths
join: → evaluate-early-access

## zimbra-auth-anomalies
<!-- Zimbra GraphQL auth anomalies -->
Detect rate-limit bypass and credential stuffing on Zimbra Network Edition GraphQL endpoints.

```sqlite target=identity role=baseline params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
baseline:
  compare: first_seen
  window: '{{lookback_days}}d'
expected: High volumes of auth attempts on GraphQL from external IPs indicate a rate-limit
  bypass attempt.
prevalence:
  by: device_hostname
  key:
  - src_endpoint_ip
  rare_below: 5
reads:
- device_hostname
- src_endpoint_ip
- request_uri
- event_type
- time
silence: not_evidence_of_absence
source: hb_auth_signin
verified: dry-run
verified_at: '2026-11-21'
~~~
SELECT device_hostname, src_endpoint_ip, request_uri, COUNT(*) AS attempts FROM hb_auth_signin WHERE (LOWER(request_uri) LIKE '%/graphql%' OR LOWER(event_type) LIKE '%soap%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days') GROUP BY device_hostname, src_endpoint_ip, request_uri HAVING attempts > 20
```

## appliance-exploit-request-paths
<!-- Exploit request paths -->
Find web requests targeting known vulnerable handlers for Zimbra, Cisco SEG, and BeyondTrust.

```sqlite target=web role=detection-candidate params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
expected: Requests to delegated sending handlers or proprietary appliance APIs from
  unauthenticated sources.
reads:
- device_hostname
- src_endpoint_ip
- url_path
- url_query
- status_code
- time
silence: not_evidence_of_absence
source: hb_http_activity
verified: dry-run
verified_at: '2026-11-21'
~~~
SELECT device_hostname, src_endpoint_ip, url_path, url_query, status_code FROM hb_http_activity WHERE (LOWER(url_path) LIKE '%savedraftrequest%' OR LOWER(url_path) LIKE '%/mrs/%' OR LOWER(url_path) LIKE '%/ecs/%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```

## evaluate-early-access
<!-- Evaluate early access patterns -->
```agent target=hunter
cite: required
context:
- scoping-vulnerable-appliances
- zimbra-auth-anomalies
- appliance-exploit-request-paths
max_iterations: 3
objective: Determine if any host shows evidence of successful exploit payloads or
  credential stuffing via bypassed rate limits.
success_criteria: A verdict of malicious | suspicious | benign per appliance host.
tools:
- endpoint
- identity
- web
```

## parallel-follow-on
<!-- Hunt for follow-on persistence and execution -->
parallel:
- → cisco-seg-script-overwrite
- → cron-driven-script-execution
join: → evaluate-compromise-chain

## cisco-seg-script-overwrite
<!-- Cisco SEG script overwrites -->
Identify the overwriting of core_watch.sh as reported in Cisco SEG CVE-2026-76519 exploitation.

```sqlite target=endpoint role=triage params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
expected: Creation or modification of the core_watch.sh script, especially if originating
  from a web-server or scanning process.
reads:
- device_hostname
- file_path
- process_name
- actor_user_name
- time
- activity_id
silence: not_evidence_of_absence
source: hb_file_activity
verified: dry-run
verified_at: '2026-11-21'
~~~
SELECT device_hostname, file_path, process_name, actor_user_name, time FROM hb_file_activity WHERE LOWER(file_path) LIKE '%/core_watch.sh' AND activity_id IN (1, 3, 5) AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```

## cron-driven-script-execution
<!-- Cron-driven script execution -->
Verify if the overwritten script is being executed with root privileges by the system cron daemon.

```sqlite target=endpoint role=enrichment params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
expected: Execution of core_watch.sh by cron verifies successful persistence and elevation
  to root.
reads:
- device_hostname
- process_path
- parent_process_name
- process_cmd_line
- user_name
- time
silence: not_evidence_of_absence
source: hb_process_activity
verified: dry-run
verified_at: '2026-11-21'
~~~
SELECT device_hostname, process_path, parent_process_name, process_cmd_line, user_name, time FROM hb_process_activity WHERE (LOWER(process_path) LIKE '%/core_watch.sh' OR LOWER(process_cmd_line) LIKE '%core_watch.sh%') AND (LOWER(parent_process_name) LIKE '%cron%' OR LOWER(parent_process_name) LIKE '%crond%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```

## evaluate-compromise-chain
<!-- Evaluate full compromise chain -->
```agent target=hunter
cite: required
context:
- evaluate-early-access
- cisco-seg-script-overwrite
- cron-driven-script-execution
max_iterations: 4
objective: Identify hosts that exhibit both initial access indicators and follow-on
  persistence or execution, confirming a successful breach.
success_criteria: A final verdict of compromised | likely_compromised | benign per
  host.
tools:
- endpoint
- identity
- web
```

## decision-route-remediation
<!-- Route remediation -->
if~: "the evaluate-compromise-chain verdict is compromised or likely_compromised for any host" (confidence: high, judge=hunter)
then: → action-isolate-compromised-host
indeterminate: → task-forensic-mailbox-audit
unavailable: → task-forensic-mailbox-audit (blind_spot: missing-appliance-access-logs)
else: → task-patch-and-harden

## action-isolate-compromised-host
<!-- Isolate compromised host -->
```action target=endpoint
~~~yaml
approval: required
~~~
Isolate the appliance from the network. Revoke all active sessions for the appliance service accounts and any administrator credentials used to manage it.
```
→ task-forensic-mailbox-audit

## task-forensic-mailbox-audit
<!-- Forensic mailbox audit -->
```manual target=analyst
Inspect Zimbra audit logs for unauthorized SaveDraftRequest events to identify which accounts were used for impersonation. Review Cisco SEG logs for TAR/PAX file processing errors that may indicate exploitation attempts.
```
→ task-patch-and-harden

## task-patch-and-harden
<!-- Patch and harden infrastructure -->
```manual target=analyst
Apply the latest security patches for Zimbra (10.1.17+), Cisco Secure Email Gateway, and BeyondTrust. Restrict access to GraphQL and SOAP endpoints to known internal administrative IP ranges.
```
→ end

Run this hunt against your own data.

In Huntbase, Scout runs every step across the tools you already connect, checks back through your history and returns a verdict with the evidence. The hunt.md file is open, so you can also take it anywhere that reads the format.

Drafted by machine (huntbase-hunt-generation, hb_google/gemini-3-flash-preview), checked by dry-run, lint, critic, then reviewed by a person.

Content pack

Sign up to open

Pack contents open in Huntbase.