A STUNning Disguise: Cling IoT Botnet Exploits Realtek SDK to Conceal C2 in Google STUN Traffic

🛡️ Verified Threat IntelDigitalSpying Research Desk
📅 October 6, 2026⏱️ 6 min read

Published on DigitalSpying.com | October 6, 2026

Author: DigitalSpying Threat Research Team

Threat Classification: IoT Botnet Evolution | Protocool Masquerading | Covert C2 Channels

|---|

Executive Summary

For nearly a decade, the threat architecture of Internet of Things (IoT) botnets remained largely predictable: automated Telnet brute-forcing or known command injection vulnerabilities paired with crude TCP, UDP, or IRC command-and-control (C2) channels. However, a newly identified botnet strain has elevated IoT evasion to an unprecedented level of sophistication. In research published by Nozomi Networks Labs and detailed by Cybersecurity News and Aviatrix Threat Research, the Cling IoT botnet is actively weaponizing persistent vulnerabilities in the Realtek Jungle SDK across routers, DVRs, and smart devices.

What sets Cling apart from predecessor botnets like Mirai and Gafgyt is its revolutionary command-and-control architecture: rather than establishing suspicious outbound TCP connections, Cling encodes its bot registration and command telemetry directly within the Session Traversal Utilities for NAT (STUN) protocol. By cloaking its beacons to mimic legitimate NAT-traversal packets exchanged with Google’s public STUN servers (stun.l.google.com:19302), Cling seamlessly blends into the millions of WebRTC video streams and VoIP calls traversing enterprise perimeters daily, effectively blinding traditional intrusion detection systems.

Technical Deep-Dive: Deconstructing the Cling Infection Lifecycle

1. Ingress: Exploiting the Persistent Realtek Jungle SDK Flaws

Cling targets edge networking equipment and connected appliances powered by Realtek system-on-chips (SoCs):

* The Exploit Vector: The botnet actively scans for internet-exposed devices running vulnerable versions of the Realtek Jungle SDK (versions 2.0 through 3.4.14B), primarily weaponizing CVE-2021-35394—a critical suite of pre-authentication command injection vulnerabilities in Realtek's diagnostic web interfaces (formSysCmd, boaf).

* Rapid Payload Staging: Upon triggering command injection, an unauthenticated shell command downloads a tailored binary compiled for MIPS, ARM, or x86 architectures from a remote staging server, deletes the downloaded file from disk, and runs purely in memory to thwart basic filesystem forensics.

[Infected Edge Router / DVR]

│

│ 1. Exploits Realtek Jungle SDK (CVE-2021-35394)

▼

[Cling Malware Memory-Resident Execution]

│

│ 2. Encapsulates Bot Metadata into STUN Binding Requests

▼

[STUN Network Traffic (UDP Port 19302)]

│

│ 3. Packets Sent to Google Public STUN Servers & C2 Reflectors

▼

[Enterprise Perimeter Firewall / IDS]

│

│ 4. Firewall Evaluates Traffic -> Classified as Benign VoIP/WebRTC!

▼

[Covert C2 Infrastructure / Botmaster]

├── Extracts Encoded Payload from STUN Attributes & Transaction IDs

└── Returns Disguised STUN Binding Responses Containing Attack Orders

2. The Covert Mechanism: Weaponizing the STUN Protocol (RFC 5389 / RFC 8489)

STUN is an established standard protocol used by video conferencing platforms, VoIP phones, and WebRTC clients to discover their public IP address and port mapping behind Network Address Translation (NAT) firewalls:

* Protocol Camouflage: STUN operates primarily over UDP (default port 3478 and 19302). Because real-time multimedia services require frictionless UDP egress, corporate and consumer firewalls almost universally allow outbound STUN traffic without deep content inspection.

* Transaction ID Encoding: Cling crafts custom 20-byte STUN headers consisting of a 2-byte Message Type, a 2-byte Message Length, a 4-byte Magic Cookie (0x2112A442), and a 12-byte Transaction ID. Cling's developers repurposed the 96-bit Transaction ID field into a cryptographic steganographic channel, encoding bot identifiers, architecture types, and telemetry.

* Abusing Google Public Infrastructure: When registering with its botmaster, Cling transmits genuine STUN Binding Requests to Google's public servers (stun.l.google.com:19302) to dynamically determine its own external NAT mapping, then uses tailored C2 nodes that listen on port 19302. To a network security analyst inspecting netflow logs, the bot’s communication appears indistinguishable from an employee participating in a Google Meet session.

3. Botnet Weaponry & Operational Capabilities

Once synchronized with its controller, Cling maintains modular offensive capabilities:

* Volumetric DDoS Flooding: Capable of launching coordinated UDP, TCP SYN, and HTTP flood attacks against critical targets.

* Residential Proxy Networks: Cling can transform infected routers into high-bandwidth SOCKS5 residential proxy nodes, leasing compromised bandwidth to cybercrime syndicates for credential stuffing and ad fraud.

* Subnet Worm Propagation: The malware inspects local routing tables (/proc/net/route) to discover adjacent private subnets, automatically scanning local network devices (such as industrial control bridges and network-attached storage) to expand its footprint behind enterprise firewalls.

Industry Impact: The Rise of "Living-Off-the-Protocol" Malware

Cling marks a pivotal transition in the threat landscape of connected devices:

* Bypassing Signature-Based Defenses: Traditional network security appliances rely heavily on port-based rules and IP reputation lists. By riding on ubiquitous, trusted infrastructure (Google STUN) and using standardized UDP protocols, Cling renders signature-based IDS sensors largely ineffective.

* The Long Tail of Firmware Supply Chains: Even years after patches are released by chip manufacturers like Realtek, white-label router and camera OEMs rarely push over-the-air firmware updates to end users, leaving tens of millions of devices permanently exposed to automated exploitation.

Defensive Blueprint & Hardening Recommendations

Network engineers, IoT device operators, and enterprise security teams must adopt behavioral and architectural mitigations:

1. Enforce Segmentation on IoT and OT VLANs

* Never place IP cameras, DVRs, building automation controllers, or consumer routers on the same VLAN as workstations or corporate servers.

* Place all connected devices inside strictly segregated IoT network segments.

2. Restrict UDP Egress from Non-Communication Subnets

* Block outbound UDP traffic on ports 3478 and 19302 from subnets that do not host authorized VoIP PBXs or approved video conferencing clients.

* If a security camera or residential router is generating persistent UDP port 19302 traffic to public IP addresses, flag the behavior as an immediate Indicator of Compromise (IoC).

3. Firmware Auditing & Remediation

* Review hardware inventories for legacy Realtek-based routers and cameras.

* Update device firmware to vendor-patched builds that incorporate updated Realtek SDK libraries.

* If an edge device has reached End-of-Life (EoL) and no security firmware is available, decommission the hardware immediately.

4. Deep Packet Inspection (DPI) of STUN Headers

* Configure Next-Generation Firewalls (NGFWs) to perform protocol anomaly checks on STUN packets, verifying that Transaction IDs follow standard RFC random distributions rather than structured, recurring cryptographic cipher patterns.

References & Threat Intelligence

* Nozomi Networks Labs Research: A STUNning Disguise — Cling Malware

* Cybersecurity News: Cling Malware Masquerades as Google STUN Traffic

* Aviatrix Threat Research Center: Realtek Jungle SDK Exploited to Deploy Cling Botnet

* CyberPress: Cling IoT Malware Masquerades as Google STUN Traffic

Classification:Cyber Security IntelligenceZero-Day AnalysisDefensive Engineering
🛡️

About the DigitalSpying Research Desk

The DigitalSpying Threat Intelligence Desk is composed of seasoned security researchers, reverse engineers, and blue team architects. Our mission is to publish reproducible, peer-audited threat analyses, hardware security evaluations, and defensive countermeasures.