The TGR-STA-0049 QUIC RAT campaign gives defenders a specific visibility problem: two custom remote access tools used HTTP/3 over UDP for command and control. Unit 42 says the espionage activity targeted Southeast Asian government entities from at least July 2025. Its report, published on 16 November 2026, describes PerfectPioneer RAT and its later modular successor, KUKURAT.
The names require care. Unit 42 tracks the activity as TGR-STA-0049 and associates it with the names REF7707, Earth Alux, and Ink Dragon. It describes Squidoor, also called FINALDRAFT, as an earlier backdoor in associated activity. The new report does not make Squidoor another name for either RAT.
TGR-STA-0049 QUIC RAT Campaign: The Attack Sequence
Unit 42 observed two routes into target environments. Exploitation of unpatched Exchange and web servers was the primary route in its investigation, followed by web shell deployment. Phishing was a second route, using decoy documents and lookalike sites to persuade users to run malware. The report does not identify a single vulnerability behind every server compromise.
Once inside, the actor used PerfectPioneer for reconnaissance and remote control. Unit 42 observed that family from July 2025 and identified six variants. By March 2026, it observed KUKURAT in networks already compromised by PerfectPioneer. This sequence shows a change in tooling inside the investigated campaign, rather than proof that every affected network hosted both families.
- EntryExposed servers or targeted phishing established access.
- FootholdWeb shells and PerfectPioneer supported reconnaissance.
- Control channelHTTP/3 over QUIC carried RAT communication.
- ExpansionStolen credentials and domain log-on scripts spread access.
- CollectionThe actor sought documents and Exchange mailbox access.
- ExfiltrationUnit 42 observed attempts to send collected material to command infrastructure.
After gaining privileged access, the actor used several credential collection methods, including LSASS memory dumping and directory service replication. Unit 42 also observed malicious domain log-on scripts after domain controller compromise. A user login could then trigger KUKURAT deployment on that endpoint. The researchers additionally observed attempts to obtain documents and broad Exchange mailbox access.
Why QUIC Changed the Detection Problem
HTTP/3 uses QUIC over UDP, while many older web inspection policies concentrate on TCP. Unit 42 says both RAT families used HTTP/3 as their primary command channel in this campaign. The reported PerfectPioneer variant could fall back to HTTP/2 when that channel failed. KUKURAT appeared in both HTTP/2 and HTTP/3 variants.
The protocol choice creates a gap when outbound UDP traffic receives less logging or inspection than TCP web traffic. Unit 42 warns that some firewalls and intrusion detection systems may miss this traffic under such policies. It does not claim that all QUIC traffic is malicious or that HTTP/3 alone identifies these RATs. Defenders need process, identity, and network evidence together.
A useful first check is whether your edge devices record outbound UDP port 443 connections with host and process context. Compare that coverage with your HTTP/2 and TLS telemetry. Then investigate unexpected long-lived or repeated QUIC sessions from servers, especially systems with recent web shell or credential alerts. These are our triage priorities drawn from Unit 42’s observed chain.
For a team building a repeatable hunt, the cyber threat intelligence platform page explains the product context for organizing actor findings. The campaign evidence should still drive local hypotheses and searches. A guide to advanced persistent threats helps distinguish long-running actor tracking from a single malware name.
PerfectPioneer and KUKURAT Are Different Families
PerfectPioneer is Go based and embeds broad remote control functions in the analyzed variant. Unit 42 describes file operations, command execution, process control, screenshots, keylogging, and tunneling capabilities. Its six variants differ in features and network behavior. The researchers caution that altered or missing compilation timestamps prevent a reliable build-by-build chronology.
Two PerfectPioneer variants contain an operator safety check that blocks a list of conspicuous or destructive commands. This is a property of the malware, not a defensive control. Unit 42 interprets it as an effort to avoid actions that might trigger alerts. The report also describes reconnection and cleanup routines that may reduce persistent evidence on a host.
KUKURAT, written in C++, moves most capabilities into modules delivered when needed. Unit 42 found support for at least 12 named modules, but that code-level support does not show that every module ran in every incident. The researchers recovered two KUKURAT variants, one using HTTP/2 and another using HTTP/3. Its core registration reports host and user details before receiving runtime settings.
KUKURAT also has a camouflage engine that can generate benign-looking network, registry, and temporary-file activity. Unit 42 says these behaviors can be enabled separately. An accompanying watchdog appeared in multiple incidents to keep the agent running. A low-alert endpoint therefore deserves examination when identity or network records indicate the larger chain.
| Feature | PerfectPioneer | KUKURAT |
|---|---|---|
| First observed | July 2025 | March 2026 |
| Implementation | Go-based agent with embedded functions | C++ core with on-demand modules |
| Transport | HTTP/3 primary, HTTP/2 fallback in analyzed variant | Separate HTTP/2 and HTTP/3 variants |
| Evasion | Operator command safety check in two variants | Configurable activity camouflage |
These distinctions matter during investigation. A rule built only for PerfectPioneer’s embedded functions may miss KUKURAT’s smaller core. A hunt restricted to HTTP/3 could miss KUKURAT’s HTTP/2 variant or PerfectPioneer’s fallback. The families belong in one campaign assessment but need separate endpoint detections.
Where to Hunt Along the Chain
Start with the initial foothold. Review exposed Exchange and web servers for unexpected processes, web shell evidence, and changes near suspicious logins. Treat suspicious document execution on user endpoints as a parallel route, since Unit 42 also saw phishing. Correlate either route with subsequent outbound connections and privilege changes.
Next, inspect identity and domain administration records. Unit 42 reports domain controller compromise followed by unauthorized log-on script changes. Audit SYSVOL and Group Policy objects for modifications, their authors, and the affected users. Correlate new script execution with endpoint process trees and KUKURAT-like network activity.
Unit 42 observed LSASS memory dumping and directory service replication in this campaign. Watch for unusual access to LSASS and replication privileges used outside normal domain controller activity. Its mitigation section also recommends behavioral monitoring of suspicious process interaction with Windows credential components. Investigate the account, host, timing, and preceding access path before concluding an alert belongs to this actor.
Review Exchange permissions as well as server compromise evidence. The researchers saw checks for ApplicationImpersonation rights and commands that granted broad mailbox access to an account. An audit should identify newly assigned impersonation roles and unusually wide FullAccess grants. Follow that review with mailbox access and document transfer records to estimate what the local evidence can establish.
Unit 42 says the targeted documents included office files and images whose names suggested international relations or political affairs. The report describes attempted exfiltration; that wording does not establish a completed transfer for each target. Preserve transfer logs and mailbox audit records before changing permissions. The incident response and threat hunting page gives the broader workflow context for this kind of evidence review.
Defensive Priorities for Government Networks
The first priority is visibility across both UDP and TCP egress. Unit 42 recommends logging and inspecting QUIC or restricting outbound UDP port 443. If an organization chooses to block it, test legitimate HTTP/3 applications and confirm that traffic falls back to inspected channels. Blocking one transport alone does not remove KUKURAT’s HTTP/2 option.
The second priority is to close the observed footholds. Inventory internet-facing Exchange and web servers, apply relevant vendor security updates, and review exposed hosts for compromise. Unit 42 does not identify one CVE or fixed version for this campaign, so remediation needs each server’s own vendor advisory. Check phishing defenses and investigate deceptive document execution as a separate entry path.
The third priority is to control lateral movement and data access. Restrict privileged accounts, review domain log-on scripts, and alert on unexpected changes to mailbox permissions. These actions follow the sequence Unit 42 documented. They also help discover persistence that can survive a simple network block.
- Map egressConfirm UDP 443 and TCP web traffic are logged with host context.
- Review footholdsExamine exposed servers and suspicious document execution.
- Audit identityCheck replication privileges, LSASS access, and domain log-on scripts.
- Review mailInspect impersonation roles and broad mailbox permissions.
- Preserve evidenceKeep network, host, and mailbox records before remediation.
- Verify controlsTest QUIC policy and confirm both transport paths remain visible.
Use these checks as a hunt plan, not a signature for attribution. HTTP/3 traffic, an administrative script, or an Exchange permission change can each have legitimate explanations. The stronger case is a sequence linking foothold, privilege use, RAT behavior, and attempted collection. Unit 42’s names describe its tracking of this activity; local evidence must support any incident conclusion.
What the Report Establishes and What Remains Local
Unit 42’s report establishes observed malware families, a campaign timeline, and several methods used in investigated environments. It also links TGR-STA-0049 with REF7707, Earth Alux, and Ink Dragon in its own tracking. It does not provide a campaign-wide victim count or a universal infection method. Neither can be inferred from the named variants or the examined cases.
For defenders, the practical test is whether existing controls can follow the whole chain. A UDP visibility check is useful, but its result needs endpoint and identity context. A KUKURAT-like process warrants scrutiny, but the report’s modular design means missing a known module is not exculpatory. Close each investigation with the evidence collected from your own systems.
Frequently Asked Questions
Are PerfectPioneer and KUKURAT the same malware?
No. Unit 42 describes PerfectPioneer as a Go-based RAT and KUKURAT as a later C++ modular family. They appeared in associated activity.
Does blocking QUIC stop this campaign?
Blocking outbound UDP 443 can force some HTTP/3 traffic onto inspected channels. KUKURAT also has an HTTP/2 variant, so endpoint and identity monitoring remain necessary.
What should an Exchange administrator inspect?
Review server compromise evidence, ApplicationImpersonation roles, and unusually broad FullAccess mailbox grants. Preserve audit records before changing permissions.
