BeyondTrust Remote Support Unauthenticated Exploitation and RCE
High severity TLP:CLEAR Research by Rapid7 Published 18 Nov 2026
Hypothesis
An intruder has exploited CVE-2026-102764 on a BeyondTrust Remote Support appliance by forging internal message origins to upload and execute a PHP payload.
Brief
Why this hunt
As detailed by Rapid7 Analysis: Unauthenticated File Upload to RCE in BeyondTrust Remote Support (CVE-2026-102764), a critical authentication bypass exists in BeyondTrust Remote Support appliances. The vulnerability allows an unauthenticated attacker to forge internal Message Routing System (MRS) origins. By convincing the system that a message originated from an internal component, an attacker accesses sensitive file-transfer services. These services permit the upload of arbitrary PHP files, leading to full remote code execution. Because BeyondTrust appliances often sit on network boundaries and possess high privileges for remote assistance, they are high-value targets for initial access and lateral movement. This hunt provides a structured way to identify if an appliance has been targeted and if those attempts were successful.
How the hunt flows
The hunt begins by scoping the environment to identify all BeyondTrust Remote Support systems. It queries the software inventory for packages matching the product name. This step ensures that subsequent high-volume log queries focus only on the relevant appliances, which minimizes performance impact and reduces noise from unrelated web servers.
The first detection phase focuses on the early stages of the exploitation chain. We search the appliance syslog for the error message DISCARDING NONLOCAL MESSAGE. This specific log entry occurs when the thin-scc worker detects an MRS message claiming to be from an internal router ID but arriving from an external connection. Simultaneously, the hunt looks for unauthenticated session starts and requests to extensionless URL paths. Attackers often use these paths to trigger their uploaded payloads, and identifying them early allows an analyst to prioritize hosts for deeper inspection.
The second phase searches for the consequences of a successful bypass. We look for file system activity involving the creation of temporary files or PHP scripts within the sdcust_thin_client directory. These files represent the staged and final payloads. To confirm execution, the hunt examines process activity for shell commands like sh or bash, or Python execution, spawned directly by the site-worker processes. We pay close attention to activity under the msfuser account, which is the default context for these services.
What the hunt cannot see
This hunt relies on the presence of local logs. If a sophisticated attacker gains root privileges and clears the syslog or web access logs, the evidence of the initial origin forgery and session start will be lost. Furthermore, the hunt cannot inspect the encrypted MRS protocol stream directly to see the specific forged IDs or the exact file paths requested before they hit the disk. Organizations should supplement this hunt with immutable off-box logging and network-level inspection for the best results.
Steps
-
Query, scoping
Identify BeyondTrust Remote Support appliances
Scope the hunt to systems running the affected software package.
sql, reads hb_software_inventorySELECT device_hostname, package_name, package_versionFROM hb_software_inventoryWHERE (LOWER(package_name) LIKE '%beyondtrust%' OR LOWER(package_name) LIKE '%remote support%')What a hit looks like. A list of hostnames belonging to Remote Support appliances. Silence means no appliances are under management or the package name differs.
-
Query, detection candidate
Origin forgery log traces
Detect the specific DISCARDING NONLOCAL MESSAGE error produced when the thin-scc worker rejects forged internal messages.
sql, reads hb_syslogSELECT device_hostname, message, timeFROM hb_syslogWHERE LOWER(message) LIKE '%discarding nonlocal message%' AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. Direct evidence of the origin forgery bypass attempt. Hits here are high-confidence indicators of exploitation.
-
Query, baseline
Rare session and menu endpoints
Find session start activity and menu interactions combined with rare, extensionless URL paths that may identify the uploaded trigger.
sql, reads hb_http_activitySELECT url_path, COUNT(DISTINCT device_hostname) AS hosts, GROUP_CONCAT(DISTINCT src_endpoint_ip) AS clients, MIN(time) AS first_seenFROM hb_http_activityWHERE (url_path = '/api/start_session' OR LOWER(url_path) LIKE '%issue-menu%' OR (instr(url_path, '.') = 0 AND length(url_path) > 5)) AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')GROUP BY url_pathHAVING hosts <= 3What a hit looks like. A rare path like /qdgusthrmlirprls that appears on few appliances across the estate shortly after a session start.
-
Agent triage
Early stage triage
Assess whether session start activity and origin errors together indicate a likely exploitation attempt.
-
Query, enrichment
Payload file deployment
Detect the creation of temporary .n$t files or PHP files in worker directories associated with the exploitation chain.
sql, reads hb_file_activitySELECT device_hostname, file_path, process_name, timeFROM hb_file_activityWHERE (LOWER(file_path) LIKE '%.php.n$t' OR LOWER(file_path) LIKE '%/sdcust_thin_client/%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. Files created in temporary customer directories or ending in the .n$t transfer suffix. Silence may occur if the attacker deletes the payload immediately.
-
Query, triage
Broad shell and site-user activity
Identify any shell execution or python scripts originating from the site worker context or running as the site user (UID 1000).
sql, reads hb_process_activitySELECT device_hostname, process_cmd_line, user_name, parent_process_name, timeFROM hb_process_activityWHERE (LOWER(process_cmd_line) LIKE '%/sh %' OR LOWER(process_cmd_line) LIKE '%/bash %' OR LOWER(process_cmd_line) LIKE '%/python%') AND (LOWER(parent_process_name) LIKE '%thin-scc%' OR LOWER(user_name) = 'msfuser') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. Shell commands run by the site's primary OS user or spawned by the thin-scc worker process.
-
Agent triage
Follow-on triage and confirmation
Synthesize early indicators with file and process evidence to reach a final verdict on the intrusion.
-
Decision
Route on verdict
Contain the threat if malicious activity is confirmed.
-
Response action
Isolate appliance
Prevent further exploitation or lateral movement from the compromised appliance.
-
Analyst task
Analyst review
Perform manual validation of suspicious leads that the agent could not definitively classify.
-
Analyst task
Close out
Document findings and schedule periodic re-runs.
Coverage
4 of 4 stages covered
| Stage and ATT&CK technique | Covered | How, or why not |
|---|---|---|
| Vulnerability Identification T1190 |
Yes | scope-beyondtrust-appliances |
| Unauthenticated Session Initialization T1190 |
Yes | lead-session-starts |
| Origin Forgery and Path Discovery T1083 |
Yes | lead-origin-errors |
| Payload Deployment and RCE T1105 T1059.006 |
Yes | follow-on-file-writes, follow-on-rce-processes |
Blind spots
- Needs MRS protocol deep packet inspection. Standard logs capture the error message but not the specific MRS message content, making it difficult to determine the exact intent of an interrupted attack. It would answer: What were the specific forged router IDs and file-transfer paths requested within the MRS stream? Remediation: Deploy TLS decryption and protocol-aware inspection for appliance traffic.
- Needs immutable off-appliance logging. If logs are only stored on the appliance, a successful RCE allows the attacker to wipe the evidence of their session start and origin forgery. It would answer: Did the attacker delete access or system logs after achieving RCE? Remediation: Ensure syslog and web access logs are streamed in real-time to a centralized, immutable SIEM.
Parameters and data
Parameters
| Parameter | Type | Default | What it is |
|---|---|---|---|
lookback_days | number | 14 | Days of history to examine. |
scope_hosts | list[host] | none | Specific BeyondTrust appliances to hunt on; leave empty for all. |
Data sources
| Source | Category | Telemetry |
|---|---|---|
| Endpoint telemetry (hb_ surfaces) | endpoint | endpoint |
| Web server / proxy logs | siem | network |
Source
An open hunt.md file. It runs anywhere that reads the format.
---
analysis: A standard detection rule might fire on the origin forgery error, but this
hunt pivots between web logs (session start), system logs (origin errors), and process
activity (shell execution) to confirm the full exploitation chain and differentiate
between scanning and a successful intrusion.
blind_spots:
- id: mrs-message-payloads
owner: Network Engineering
question: What were the specific forged router IDs and file-transfer paths requested
within the MRS stream?
remediation: Deploy TLS decryption and protocol-aware inspection for appliance traffic.
requires: MRS protocol deep packet inspection
risk: Standard logs capture the error message but not the specific MRS message content,
making it difficult to determine the exact intent of an interrupted attack.
stage: origin-forgery-and-path-discovery
- id: log-deletion-on-host
owner: Security Operations
question: Did the attacker delete access or system logs after achieving RCE?
remediation: Ensure syslog and web access logs are streamed in real-time to a centralized,
immutable SIEM.
requires: immutable off-appliance logging
risk: If logs are only stored on the appliance, a successful RCE allows the attacker
to wipe the evidence of their session start and origin forgery.
stage: payload-deployment-and-rce
coverage:
- stage: vulnerability-identification
status: covered
steps:
- scope-beyondtrust-appliances
- stage: unauthenticated-session-initialization
status: covered
steps:
- lead-session-starts
- stage: origin-forgery-and-path-discovery
status: covered
steps:
- lead-origin-errors
- stage: payload-deployment-and-rce
status: covered
steps:
- follow-on-file-writes
- follow-on-rce-processes
guardrails:
claims: no_unsupported
evidence: citation_required
missing_data: not_benign
telemetry: untrusted
hunt:
applicability: campaign-specific
handoff: promote-to-detection
justification: Protecting BeyondTrust Remote Support appliances from unauthenticated
RCE is a critical priority, as these systems often hold privileged access to the
internal network. A negative result confirms that the appliances are not currently
being exploited by known tradecraft.
methodology: model-assisted
trigger: intel-report
hypothesis: An intruder has exploited CVE-2026-102764 on a BeyondTrust Remote Support
appliance by forging internal message origins to upload and execute a PHP payload.
labels:
- hunt
- attack.t1190
- attack.t1083
- attack.t1105
- attack.t1059.006
- discovery
- execution
- initial access
name: BeyondTrust Remote Support Unauthenticated Exploitation and RCE
parameters:
lookback_days:
default: '14'
description: Days of history to examine.
from:
kind: manual
observed: '2026-10-30'
ref: hunt-standard
type: number
scope_hosts:
default: []
description: Specific BeyondTrust appliances to hunt on; leave empty for all.
from:
kind: manual
observed: '2026-10-30'
ref: hunt-standard
type: list[host]
provenance:
authors:
- name: Huntbase hunt generation
org: huntbase.io
generated:
by: huntbase-hunt-generation
from: https://www.rapid7.com/blog/post/ra-rapid7-analysis-unauthenticated-file-upload-to-rce-in-beyondtrust-remote-support-cve-2026-102764
gates:
- dry-run
- lint
model: hb_google/gemini-3-flash-preview
rationale: Start by identifying all BeyondTrust Remote Support appliances via software
inventory. Focus the HTTP and syslog queries on these specific hosts to reduce noise
from general web traffic.
references:
- name: 'Rapid7 Analysis: Unauthenticated File Upload to RCE in BeyondTrust Remote
Support (CVE-2026-102764)'
url: https://www.rapid7.com/blog/post/ra-rapid7-analysis-unauthenticated-file-upload-to-rce-in-beyondtrust-remote-support-cve-2026-102764
related:
- hunt: beyondtrust-unauth-session-hijack
reason: The same unauthenticated session start can be used to attempt session hijacking
or other protocol-level attacks.
relation: sibling
scenario:
stages:
- name: Vulnerability Identification
observables:
- BeyondTrust Remote Support version 24.1.2
- Build 1524046
- Revision 4208-ddcf90715106aa0c0e19c103f8d2c412af4beff3
- CVE-2026-102764
slug: vulnerability-identification
tactic: initial-access
techniques:
- T1190
- name: Unauthenticated Session Initialization
observables:
- POST /api/start_session
- WebSocket connection to /nw
- brain.log entries for 'ingredi support desk customer' connections
- 'User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 14.7; rv:133.0) Gecko/20100101
Firefox/133.0'
slug: unauthenticated-session-initialization
tactic: initial-access
techniques:
- T1190
- name: Origin Forgery and Path Discovery
observables:
- Forged router ID in MRS message source addresses
- Access to /proc/self/environ via internal file-transfer service
- thin-scc-errors.log containing 'DISCARDING NONLOCAL MESSAGE'
- thin-scc.log entry for DownloadAnonymousFiles native provider access
slug: origin-forgery-and-path-discovery
tactic: discovery
techniques:
- T1083
- name: Payload Deployment and RCE
observables:
- Upload of PHP files ending in .php.n$t
- File paths under $BG_app_root/data/tmp/sdcust_thin_client/
- PHP execution as user msfuser (UID 1000)
- HTTP GET/POST requests to extensionless endpoints (e.g., /qdgusthrmlirprls)
in the public web directory
- 'User-Agent: Python-urllib/3.9'
slug: payload-deployment-and-rce
tactic: execution
techniques:
- T1105
- T1059.006
summary: Attackers exploit an origin-validation flaw in BeyondTrust Remote Support
(CVE-2026-102764) by forging internal message routing IDs to bypass authentication.
This allows unauthenticated remote file operations, which the attacker uses to
retrieve environment paths, upload a malicious PHP payload, and achieve remote
code execution as the site's operating-system user.
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
web:
category: siem
name: Web server / proxy logs
telemetry:
- network
tlp: clear
type: investigation
---
# BeyondTrust Remote Support Unauthenticated Exploitation and RCE
This hunt targets a critical origin-validation flaw in the BeyondTrust Remote Support appliance. Attackers can start anonymous customer sessions and forge message source addresses to reach internal file-transfer services. The hunt identifies early indicators of this forgery in appliance logs, correlates them with rare session initialization patterns, and then searches for follow-on payload deployment and process execution. Because the appliance is a closed system, these behavioral signals in the web and system logs are the primary means of detecting unauthenticated remote code execution.
## scope-beyondtrust-appliances
<!-- Identify BeyondTrust Remote Support appliances -->
Scope the hunt to systems running the affected software package.
```sqlite target=endpoint role=scoping
~~~yaml
expected: A list of hostnames belonging to Remote Support appliances. Silence means
no appliances are under management or the package name differs.
reads:
- device_hostname
- package_name
- package_version
silence: not_evidence_of_absence
source: hb_software_inventory
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT device_hostname, package_name, package_version FROM hb_software_inventory WHERE (LOWER(package_name) LIKE '%beyondtrust%' OR LOWER(package_name) LIKE '%remote support%')
```
## parallel-early-leads
<!-- Hunt for early-stage leads -->
parallel:
- → lead-origin-errors
- → lead-session-starts
join: → early-triage
## lead-origin-errors
<!-- Origin forgery log traces -->
Detect the specific DISCARDING NONLOCAL MESSAGE error produced when the thin-scc worker rejects forged internal messages.
```sqlite target=endpoint role=detection-candidate params=(scope_hosts=scope_hosts, lookback_days=lookback_days)
~~~yaml
expected: Direct evidence of the origin forgery bypass attempt. Hits here are high-confidence
indicators of exploitation.
reads:
- device_hostname
- message
- time
silence: not_evidence_of_absence
source: hb_syslog
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT device_hostname, message, time FROM hb_syslog WHERE LOWER(message) LIKE '%discarding nonlocal message%' AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```
## lead-session-starts
<!-- Rare session and menu endpoints -->
Find session start activity and menu interactions combined with rare, extensionless URL paths that may identify the uploaded trigger.
```sqlite target=web role=baseline params=(scope_hosts=scope_hosts, lookback_days=lookback_days)
~~~yaml
baseline:
compare: first_seen
window: '{{lookback_days}}d'
expected: A rare path like /qdgusthrmlirprls that appears on few appliances across
the estate shortly after a session start.
prevalence:
by: device_hostname
key:
- url_path
rare_below: 3
reads:
- url_path
- device_hostname
- src_endpoint_ip
- time
silence: not_evidence_of_absence
source: hb_http_activity
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT url_path, COUNT(DISTINCT device_hostname) AS hosts, GROUP_CONCAT(DISTINCT src_endpoint_ip) AS clients, MIN(time) AS first_seen FROM hb_http_activity WHERE (url_path = '/api/start_session' OR LOWER(url_path) LIKE '%issue-menu%' OR (instr(url_path, '.') = 0 AND length(url_path) > 5)) AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days') GROUP BY url_path HAVING hosts <= 3
```
## early-triage
<!-- Early stage triage -->
```agent target=hunter
cite: required
context:
- lead-origin-errors
- lead-session-starts
max_iterations: 4
objective: Identify hosts where unauthenticated session initialization or issue-menu
access is followed by origin forgery errors or rare extensionless URL requests.
success_criteria: A per-host verdict of suspicious if session start patterns and error
logs align temporally.
tools:
- endpoint
- web
```
## parallel-follow-on
<!-- Hunt for payload and RCE -->
parallel:
- → follow-on-file-writes
- → follow-on-rce-processes
join: → final-triage
## follow-on-file-writes
<!-- Payload file deployment -->
Detect the creation of temporary .n$t files or PHP files in worker directories associated with the exploitation chain.
```sqlite target=endpoint role=enrichment params=(scope_hosts=scope_hosts, lookback_days=lookback_days)
~~~yaml
expected: Files created in temporary customer directories or ending in the .n$t transfer
suffix. Silence may occur if the attacker deletes the payload immediately.
reads:
- device_hostname
- file_path
- process_name
- time
silence: not_evidence_of_absence
source: hb_file_activity
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT device_hostname, file_path, process_name, time FROM hb_file_activity WHERE (LOWER(file_path) LIKE '%.php.n$t' OR LOWER(file_path) LIKE '%/sdcust_thin_client/%') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```
## follow-on-rce-processes
<!-- Broad shell and site-user activity -->
Identify any shell execution or python scripts originating from the site worker context or running as the site user (UID 1000).
```sqlite target=endpoint role=triage params=(scope_hosts=scope_hosts, lookback_days=lookback_days)
~~~yaml
expected: Shell commands run by the site's primary OS user or spawned by the thin-scc
worker process.
reads:
- device_hostname
- process_cmd_line
- user_name
- parent_process_name
- time
silence: not_evidence_of_absence
source: hb_process_activity
verified: dry-run
verified_at: '2026-11-18'
~~~
SELECT device_hostname, process_cmd_line, user_name, parent_process_name, time FROM hb_process_activity WHERE (LOWER(process_cmd_line) LIKE '%/sh %' OR LOWER(process_cmd_line) LIKE '%/bash %' OR LOWER(process_cmd_line) LIKE '%/python%') AND (LOWER(parent_process_name) LIKE '%thin-scc%' OR LOWER(user_name) = 'msfuser') AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```
## final-triage
<!-- Follow-on triage and confirmation -->
```agent target=hunter
cite: required
context:
- early-triage
- follow-on-file-writes
- follow-on-rce-processes
max_iterations: 5
objective: Determine if the combination of origin errors, rare URL requests, file
deployment, and subsequent process activity establishes a successful unauthenticated
RCE intrusion.
success_criteria: A malicious verdict for any host where the timeline shows unauth
session -> origin error -> file write -> shell execution.
tools:
- endpoint
- web
```
## route-on-verdict
<!-- Route on verdict -->
if~: "The final triage verdict is malicious for at least one host based on origin errors and confirmed code execution." (confidence: high, judge=hunter)
then: → isolate-appliance
indeterminate: → analyst-review
unavailable: → analyst-review (blind_spot: log-deletion-on-host)
else: → close-out
## isolate-appliance
<!-- Isolate appliance -->
```action target=endpoint
~~~yaml
approval: required
~~~
Isolate the Remote Support appliance and revoke all active customer and representative sessions.
```
→ analyst-review
## analyst-review
<!-- Analyst review -->
```manual target=analyst
Review the syslog origin errors and HTTP extensionless URL requests; confirm if they align with the reported CVE-2026-102764 behavior.
```
→ close-out
## close-out
<!-- Close out -->
```manual target=analyst
Record the hosts examined and the presence or absence of exploitation indicators. Recommend immediate patching if vulnerable versions were found.
```
→ 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, then reviewed by a person.