Overlord RAT Exploitation and Data Encryption
High severity TLP:CLEAR Part 2 of 2 Research by Cisco Talos Published 19 Nov 2026
Hypothesis
An adversary exploits a public-facing application to deploy the Overlord RAT and then encrypts data, bypassing traditional network blocks by resolving C2 coordinates via blockchain transactions.
Brief
Why now
Adversaries increasingly use public blockchains to host command-and-control (C2) instructions, a technique Cisco Talos describes in The challenge of defending against bulletproof dead drops on public blockchains. Because these transactions occur on legitimate decentralized ledgers like Solana or Ethereum, traditional domain or IP-based blocking often fails. This hunt focuses on the endpoint footprint of the Overlord RAT, which uses these blockchain dead drops after exploiting public-facing web applications.
How the hunt flows
The hunt begins with scoping. The first query searches software inventory to identify hosts running web server packages such as IIS, Apache, or Nginx. This limits the search to the primary attack surface and reduces noise in later steps.
Next, the hunt looks for application error logs. Specifically, it searches for access violation crashes (exception code c0000005) involving the identified web server binaries. An analyst evaluates these leads to determine if the crashes suggest a directed memory corruption exploit rather than routine software instability. This step acts as a gate for more expensive telemetry searches.
If the crash lead is suspicious, the hunt fans out into two parallel queries. The first query checks process activity for processes with no backing file on disk. This identifies the Overlord RAT's decrypter, which resides only in memory. The second query baselines file activity to find any process modifying more than 50 files in a short window, signaling the transition to ransomware encryption.
Finally, the analyst synthesizes the evidence. A successful hunt correlates the initial application crash, the presence of a fileless process, and high-volume file modifications on a single host. This combined evidence confirms a successful Overlord RAT intrusion and triggers isolation protocols.
What the hunt cannot see
This hunt depends on the exploit generating a detectable memory corruption event. If an adversary uses a logical exploit or a very stable memory exploit that does not crash the service, the hunt will not find the initial access lead. Additionally, if the endpoint agent truncates high-volume file events or does not monitor specific directories, the mass encryption phase may go undetected. The hunt also stays on the endpoint; it does not monitor the specific RPC calls to blockchain nodes.
In this series
Steps
-
Query, scoping
Identify internet-facing web assets
Identify hosts running web server software to focus the hunt on the primary attack surface.
sql, reads hb_software_inventorySELECT device_hostname, package_name, package_versionFROM hb_software_inventoryWHERE (instr(',' || '{{web_server_packages}}' || ',', ',' || LOWER(package_name) || ',') > 0 OR LOWER(package_name) LIKE '%web%server%')What a hit looks like. A list of hosts running web server packages. Silence implies no known web server software was found in inventory.
-
Query, triage
Web server memory corruption lead
Find memory corruption events that likely represent exploitation attempts against web-facing services.
sql, reads hb_application_errorSELECT device_hostname, app_name, exception_code, message, timeFROM hb_application_errorWHERE exception_code = 'c0000005' AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. Access violation crashes (c0000005) in web server binaries or their modules. Silence suggests no obvious memory-corruption exploits occurred.
-
Agent triage
Evaluate exploit crash lead
Determine if observed application errors suggest a directed exploit rather than routine software instability.
-
Decision
Gate on exploit lead
Gate expensive post-exploit searches behind a verified suspicious crash signal.
-
Query, detection candidate
Fileless process activity
Detect the Overlord RAT by identifying processes that have no backing file on disk, signifying in-memory execution.
sql, reads hb_process_activitySELECT device_hostname, process_name, process_cmd_line, on_disk, timeFROM hb_process_activityWHERE on_disk = 0 AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')What a hit looks like. A process running with on_disk = 0. This is the primary indicator of the Overlord RAT's in-memory decrypter.
-
Query, baseline
Rapid file modification baseline
Identify ransomware behavior by stack-counting high-frequency file modifications per process.
sql, reads hb_file_activitySELECT device_hostname, process_name, COUNT(DISTINCT file_path) AS unique_files, MIN(time) AS first_mod, MAX(time) AS last_modFROM hb_file_activityWHERE activity_id IN (1, 3, 5) AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')GROUP BY device_hostname, process_nameHAVING unique_files > 50ORDER BY unique_files DESCWhat a hit looks like. A rare process modifying more than 50 files. Silence means no mass file changes were detected.
-
Agent triage
Synthesize exploit and impact evidence
Combine exploitation signals with follow-on execution and disk activity to reach a final verdict.
-
Decision
Route on final verdict
Direct the hunt based on the final synthesized verdict.
-
Response action
Isolate infected host
Contain the threat and stop the spread of ransomware.
-
Analyst task
Analyst forensic review
Confirm the infection details and identify the specific blockchain records used for C2 resolution.
-
Analyst task
Close out hunt
Finalize documentation and record negative findings.
Coverage
3 of 5 stages covered
| Stage and ATT&CK technique | Covered | How, or why not |
|---|---|---|
| Exploitation of Public-Facing Application T1190 |
Yes | scope-web-assets, exploit-crash-lead |
| Overlord RAT Execution | Yes | fileless-execution-rat |
| Data Encrypted for Impact T1486 |
Yes | rapid-file-encryption |
| Blockchain RPC Dead-Drop Resolution T1090.004 |
Out of scope | Belongs to another part of the 'The challenge of defending against bulletproof dead drops on public blockchains' series. |
| C2 Infrastructure Beaconing | Out of scope | Belongs to another part of the 'The challenge of defending against bulletproof dead drops on public blockchains' series. |
Blind spots
- Needs hb_application_error for all web server service accounts. A quiet exploit that achieves code execution without an access violation will bypass the gated logic. It would answer: whether the exploit succeeded without generating a crash event
- Needs hb_file_activity with low-latency collection. If the agent truncates file events or does not monitor specific directories, the encryption phase may go undetected. It would answer: whether the EDR captures the speed and volume of file modifications
Parameters and data
Parameters
| Parameter | Type | Default | What it is |
|---|---|---|---|
lookback_days | number | 14 | Days of history to examine. |
scope_hosts | list[host] | none | Optional list of hostnames to focus the hunt on; if empty, the entire estate is checked. |
web_server_packages | list[string] | iis, apache, nginx, tomcat, httpd | Common web server package names to identify the attack surface. |
Data sources
| Source | Category | Telemetry |
|---|---|---|
| Endpoint telemetry (hb_ surfaces) | endpoint | endpoint |
Source
An open hunt.md file. It runs anywhere that reads the format.
---
analysis: A single rule may detect a crash or a fileless process, but this hunt correlates
instability, in-memory execution, and rapid disk activity across three different
surfaces to provide a high-confidence forensic picture of a ransomware intrusion.
blind_spots:
- id: no-crash-telemetry
question: whether the exploit succeeded without generating a crash event
requires: hb_application_error for all web server service accounts
risk: A quiet exploit that achieves code execution without an access violation will
bypass the gated logic.
stage: initial-access-exploit
- id: no-file-telemetry
question: whether the EDR captures the speed and volume of file modifications
requires: hb_file_activity with low-latency collection
risk: If the agent truncates file events or does not monitor specific directories,
the encryption phase may go undetected.
stage: impact-data-encryption
coverage:
- stage: initial-access-exploit
status: covered
steps:
- scope-web-assets
- exploit-crash-lead
- stage: execution-overlord-rat
status: covered
steps:
- fileless-execution-rat
- stage: impact-data-encryption
status: covered
steps:
- rapid-file-encryption
- reason: Belongs to another part of the 'The challenge of defending against bulletproof
dead drops on public blockchains' series.
stage: c2-blockchain-dead-drop-resolution
status: out_of_scope
- reason: Belongs to another part of the 'The challenge of defending against bulletproof
dead drops on public blockchains' series.
stage: c2-primary-beaconing
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: Public-facing applications remain a primary intrusion vector. This
hunt addresses the risk of blockchain-backed malware that evades traditional infrastructure
blocks by focusing on the local endpoint footprint of exploitation and impact.
methodology: model-assisted
trigger: intel-report
hypothesis: An adversary exploits a public-facing application to deploy the Overlord
RAT and then encrypts data, bypassing traditional network blocks by resolving C2
coordinates via blockchain transactions.
labels:
- hunt
- attack.t1190
- attack.t1486
- command and control
- execution
- impact
- initial access
name: Overlord RAT Exploitation and Data Encryption
parameters:
lookback_days:
default: '14'
description: Days of history to examine.
from:
kind: article
observed: '2026-11-18'
ref: https://blog.talosintelligence.com/the-challenge-of-defending-against-bulletproof-dead-drops-on-public-blockchains/
type: number
scope_hosts:
default: []
description: Optional list of hostnames to focus the hunt on; if empty, the entire
estate is checked.
from:
kind: manual
observed: '2026-11-18'
ref: scoping-step
type: list[host]
web_server_packages:
default:
- iis
- apache
- nginx
- tomcat
- httpd
description: Common web server package names to identify the attack surface.
from:
kind: article
observed: '2026-11-18'
ref: https://blog.talosintelligence.com/the-challenge-of-defending-against-bulletproof-dead-drops-on-public-blockchains/
type: list[string]
provenance:
authors:
- name: Huntbase hunt generation
org: huntbase.io
generated:
by: huntbase-hunt-generation
from: https://blog.talosintelligence.com/the-challenge-of-defending-against-bulletproof-dead-drops-on-public-blockchains/
gates:
- dry-run
- lint
- critic
model: hb_google/gemini-3-flash-preview
rationale: The hunt begins by identifying web server infrastructure; ensure the web_server_packages
parameter is tailored to your environment.
references:
- name: "Cisco Talos \u2014 The challenge of defending against bulletproof dead drops\
\ on public blockchains"
url: https://blog.talosintelligence.com/the-challenge-of-defending-against-bulletproof-dead-drops-on-public-blockchains/
related:
- hunt: c2-blockchain-dead-drop-resolution
reason: That hunt focuses on outbound network RPC calls to blockchain nodes, while
this hunt focuses on the endpoint execution and resulting impact.
relation: out-of-scope-alternative
- hunt: decentralized-blockchain-c2-resolution
relation: follows
scenario:
stages:
- name: Exploitation of Public-Facing Application
observables:
- anomalous web server errors
- unexpected inbound traffic to public-facing web applications
slug: initial-access-exploit
tactic: initial-access
techniques:
- T1190
- name: Overlord RAT Execution
observables:
- Overlord RAT binary
- in-memory decryption of C2 configurations
- lightweight decrypter process activity
slug: execution-overlord-rat
tactic: execution
- name: Blockchain RPC Dead-Drop Resolution
observables:
- outbound JSON-RPC queries to Infura (infura.io)
- outbound JSON-RPC queries to QuickNode (quicnode.pro)
- outbound JSON-RPC queries to Tatum (tatum.io)
- 'JSON-RPC methods: eth_call, getTransaction, getSignaturesForAddress'
- base-encoded data in HTTP POST request bodies
- DNS resolution for public blockchain node providers
slug: c2-blockchain-dead-drop-resolution
tactic: command-and-control
techniques:
- T1090.004
- name: C2 Infrastructure Beaconing
observables:
- outbound connections to dynamically resolved C2 IP addresses
- beaconing traffic to previously unknown domains
slug: c2-primary-beaconing
tactic: command-and-control
- name: Data Encrypted for Impact
observables:
- rapid file modification and encryption
- ransomware notes on system drives
slug: impact-data-encryption
tactic: impact
techniques:
- T1486
summary: The threat actor UAT-12643 utilizes the Overlord RAT to retrieve encrypted
command-and-control configurations from public blockchains including Bitcoin,
Ethereum, and Solana. By using JSON-RPC queries to public node providers like
Infura and QuickNode, the malware resolves its C2 infrastructure through immutable
ledger data such as Solana Memo transactions or Ethereum smart contracts.
series:
index: 2
slug: the-challenge-of-defending-against-bulletproof-dead-drops-on-public-blockchains
title: The challenge of defending against bulletproof dead drops on public blockchains
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
tlp: clear
type: investigation
---
# Overlord RAT Exploitation and Data Encryption
By correlating application instability with subsequent fileless activity and mass encryption, the hunt identifies the full intrusion lifecycle. The adversary exploits internet-facing web assets to achieve initial access, often leaving a trail of memory-corruption crashes. The hunt first isolates these crash leads before deploying a fan-out of queries to find fileless execution and rapid file modifications. Finally, the analyst confirms the presence of the Overlord RAT by weighing the combined evidence from application errors, endpoint process state, and disk activity.
## scope-web-assets
<!-- Identify internet-facing web assets -->
Identify hosts running web server software to focus the hunt on the primary attack surface.
```sqlite target=endpoint role=scoping params=(web_server_packages=web_server_packages)
~~~yaml
expected: A list of hosts running web server packages. Silence implies no known web
server software was found in inventory.
reads:
- device_hostname
- package_name
- package_version
silence: not_evidence_of_absence
source: hb_software_inventory
verified: dry-run
verified_at: '2026-11-19'
~~~
SELECT device_hostname, package_name, package_version FROM hb_software_inventory WHERE (instr(',' || '{{web_server_packages}}' || ',', ',' || LOWER(package_name) || ',') > 0 OR LOWER(package_name) LIKE '%web%server%')
```
## exploit-crash-lead
<!-- Web server memory corruption lead -->
Find memory corruption events that likely represent exploitation attempts against web-facing services.
```sqlite target=endpoint role=triage params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
expected: Access violation crashes (c0000005) in web server binaries or their modules.
Silence suggests no obvious memory-corruption exploits occurred.
reads:
- device_hostname
- app_name
- exception_code
- message
- time
silence: not_evidence_of_absence
source: hb_application_error
verified: dry-run
verified_at: '2026-11-19'
~~~
SELECT device_hostname, app_name, exception_code, message, time FROM hb_application_error WHERE exception_code = 'c0000005' AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```
## evaluate-lead-agent
<!-- Evaluate exploit crash lead -->
```agent target=hunter
cite: required
context:
- exploit-crash-lead
max_iterations: 3
objective: Determine whether the crashes in exploit-crash-lead suggest a directed
memory corruption exploit against a web application.
success_criteria: A per-host verdict of suspicious | benign citing the crashed application
and frequency.
tools:
- endpoint
```
## gate-decision
<!-- Gate on exploit lead -->
if~: "the evaluate-lead-agent verdict is suspicious for at least one host" (confidence: medium, judge=hunter)
then: → parallel-fan-out
indeterminate: → close-out
unavailable: → close-out (blind_spot: no-crash-telemetry)
else: → close-out
## parallel-fan-out
<!-- Fan-out to execution and impact -->
parallel:
- → fileless-execution-rat
- → rapid-file-encryption
join: → triage-agent
## fileless-execution-rat
<!-- Fileless process activity -->
Detect the Overlord RAT by identifying processes that have no backing file on disk, signifying in-memory execution.
```sqlite target=endpoint role=detection-candidate params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
expected: A process running with on_disk = 0. This is the primary indicator of the
Overlord RAT's in-memory decrypter.
reads:
- device_hostname
- process_name
- process_cmd_line
- on_disk
- time
silence: not_evidence_of_absence
source: hb_process_activity
verified: dry-run
verified_at: '2026-11-19'
~~~
SELECT device_hostname, process_name, process_cmd_line, on_disk, time FROM hb_process_activity WHERE on_disk = 0 AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days')
```
## rapid-file-encryption
<!-- Rapid file modification baseline -->
Identify ransomware behavior by stack-counting high-frequency file modifications per process.
```sqlite target=endpoint role=baseline params=(lookback_days=lookback_days, scope_hosts=scope_hosts)
~~~yaml
baseline:
compare: first_seen
window: '{{lookback_days}}d'
expected: A rare process modifying more than 50 files. Silence means no mass file
changes were detected.
prevalence:
by: device_hostname
key:
- process_name
rare_below: 3
reads:
- device_hostname
- process_name
- activity_id
- file_path
- time
silence: not_evidence_of_absence
source: hb_file_activity
verified: dry-run
verified_at: '2026-11-19'
~~~
SELECT device_hostname, process_name, COUNT(DISTINCT file_path) AS unique_files, MIN(time) AS first_mod, MAX(time) AS last_mod FROM hb_file_activity WHERE activity_id IN (1, 3, 5) AND ('{{scope_hosts}}' = '' OR instr(',' || '{{scope_hosts}}' || ',', ',' || device_hostname || ',') > 0) AND time >= datetime('now', '-{{lookback_days}} days') GROUP BY device_hostname, process_name HAVING unique_files > 50 ORDER BY unique_files DESC
```
## triage-agent
<!-- Synthesize exploit and impact evidence -->
```agent target=hunter
cite: required
context:
- evaluate-lead-agent
- fileless-execution-rat
- rapid-file-encryption
max_iterations: 6
objective: Determine if the suspicious crash, fileless execution, and high-volume
file modification together indicate a successful Overlord RAT intrusion and ransomware
event.
success_criteria: A per-host verdict of malicious | suspicious | benign citing evidence
from all stages.
tools:
- endpoint
```
## route-decision
<!-- Route on final verdict -->
if~: "the triage-agent verdict is malicious for at least one host" (confidence: high, judge=hunter)
then: → isolate-infected-host
indeterminate: → analyst-review
unavailable: → analyst-review (blind_spot: no-file-telemetry)
else: → close-out
## isolate-infected-host
<!-- Isolate infected host -->
```action target=endpoint
~~~yaml
approval: required
~~~
Isolate the host from the network. Capture process memory for the fileless process before remediation to identify the C2 resolved via blockchain.
```
→ analyst-review
## analyst-review
<!-- Analyst forensic review -->
```manual target=analyst
Review the file modification paths to identify the ransomware extension. Examine the isolated process memory to find the Solana or Ethereum RPC calls used to retrieve the C2 coordinates.
```
→ end
## close-out
<!-- Close out hunt -->
```manual target=analyst
Record the hosts examined. If no suspicious activity was found, update the web server scoping lists for future hunts.
```
→ 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.