Ace Security Desk – Hackers Exploit Unpatched Microsoft Exchange Servers to Deploy Custom RATs Across Southeast Asia by Cybersecurity’s Tushar Subhra Dutta Published: Nov. 17, 2026

Hackers are exploiting unpatched Microsoft Exchange servers and vulnerable web servers to break into government networks across Southeast Asia, deploying custom remote access trojans built for quiet, long-term spying.
The campaign combines server exploitation, targeted phishing, credential theft, and mailbox access to pursue sensitive government information.
The activity has targeted government entities since at least July 2025. Investigators observed attackers deploying PerfectPioneer RAT before introducing its modular successor, KUKURAT, in March 2026.
The operation focused on maintaining access and collecting intelligence, with targeted filenames suggesting an interest in international relations and internal political affairs.
Researchers from Unit 42 identified both previously undocumented malware families and tracked the campaign as TGR-STA-0049, also known as REF7707, Earth Alux, and Ink Dragon.
The investigation shows how the attackers steadily updated their tools while combining familiar intrusion methods with newer ways to hide network traffic and endpoint activity.
Hackers Exploit Unpatched Microsoft Exchange Servers
Server exploitation was the primary entry method. Attackers compromised vulnerable Exchange and web servers, planted web shells, and used that access to run reconnaissance commands and deploy malware.
The report does not identify the specific Exchange vulnerabilities exploited, so the intrusions should not be attributed to a particular CVE or exploit chain.
The approach echoes earlier Exchange server backdoor attacks involving other threat groups, although those incidents do not establish a connection to this campaign.
.webp)
Here, researchers observed a mixture of custom and publicly available web shells, including ASPXSpy, Godzilla, and Suo5 variants. Phishing provided a second route into targeted systems.
Attackers used lookalike government domains and executable files disguised with PDF icons. Opening these files launched malware alongside a decoy document.
One lure impersonated an official Myanmar election notice, while another delivery method ran a genuine GlobalProtect installer as malware executed in the background.
PerfectPioneer is written in Go, and researchers recovered six variants. The most complete version supports file transfers, shell commands, process management, screenshots, keylogging, code injection, and SOCKS5 proxying.
Some variants also contain an operator safety mechanism that blocks destructive or conspicuous commands, reducing the chance that an attacker’s mistake exposes the intrusion.
Both malware families support HTTP/3 communications over QUIC, which uses UDP rather than TCP. This matters because some network defenses inspect TCP traffic more closely than UDP traffic.
PerfectPioneer’s analyzed communication framework can fall back to HTTP/2, while KUKURAT has separate HTTP/2 and HTTP/3 variants. Blocking QUIC alone therefore does not remove the threat.
KUKURAT Hides Activity While Attackers Steal Credentials and Email
KUKURAT shifts most capabilities into separate DLL modules retrieved from the command server when needed. Written in C++, its code supports at least 12 named modules covering functions such as screenshots, keylogging, file management, scanning, remote execution, and lateral movement. This separates the small core agent from much of its operational functionality.
Unlike the distinct TinyRCT government espionage campaign, this operation also features a coordinated camouflage engine. KUKURAT can generate harmless-looking web requests, imitate browser-related registry activity, and rapidly create and delete temporary files.
.webp)
These actions aim to bury meaningful signals among artificial noise rather than simply conceal a single malicious file. Attackers also deployed a watchdog component that could restart KUKURAT after it stopped responding.
After compromising domain controllers, they modified domain log-on scripts so legitimate user authentication triggered malware downloads and execution on additional endpoints. Researchers also observed the use of established remote administration methods for movement inside compromised networks.
Credential theft included LSASS memory dumping, DCSync, Mimikatz, ProcDump, and SharpDPAPI. The attackers then examined Exchange impersonation rights and, in observed cases, granted a compromised account broad mailbox access.
Their document collection targeted office files, PDFs, and images, linking endpoint compromise directly to the campaign’s intelligence-gathering objective.
The pattern reinforces lessons from other Exchange server exploitation campaigns without implying shared attribution. Defenders should patch exposed servers, investigate existing compromises, audit domain log-on scripts, and review Exchange permissions.
Unit 42 additionally recommends inspecting or restricting QUIC traffic and monitoring suspicious credential-dumping behavior.
Removing malware without addressing stolen credentials, mailbox permissions, and domain persistence may leave attackers with routes back into the environment.
At Sterling Publishing & Media Service Agency, we prioritise transparency and accountability in all our operations. We wish to clarify that we are not responsible for any external content, hyperlinks, or costs associated with our services. Nevertheless, we remain committed to delivering outstanding services and greatly value your continued support. Thank you for your trust in us.