A cyberespionage group that has spent more than a year inside Southeast Asian government networks is talking to its malware over HTTP/3, the newest version of the web's core protocol, in a bid to slip past network defenses that were built for older traffic. Researchers at Palo Alto Networks' Unit 42 detailed the campaign in a report by Tom Fakterman published on November 16, 2026.
Unit 42 tracks the activity as TGR-STA-0049, a cluster also known as REF7707, Earth Alux and Ink Dragon, which IntelFusions profiles as Jewelbug. According to the report, this campaign has targeted Southeast Asian government entities since at least July 2025, while the group itself has been active since at least March 2023 against government organizations in Southeast Asia, South America and Europe. It was previously seen deploying Squidoor, also known as FINALDRAFT, a backdoor for Windows and Linux. We last covered this cluster in August, when one of its control panels was tied to both spying and crypto fraud.
Two new backdoors, nine months of upgrades
The campaign introduced two previously undocumented remote access trojans (RATs), malware that hands an operator remote control of a machine. PerfectPioneer RAT, written in Go, was first spotted in July 2025, and Unit 42 recovered six variants of it. The most complete variant, which its author labelled newhttp3, can browse and move files, run shell commands, start and kill processes, inject code into other processes, log keystrokes, take screenshots and act as a SOCKS5 proxy.
In March 2026 the operators began rolling out its successor, KUKURAT, on networks PerfectPioneer had already compromised. KUKURAT is written in C++ and is fully modular: almost every capability lives in a separate DLL that the command server pushes into memory only when needed, and its code supports at least 12 named modules. Because the core binary carries no advanced feature code, Unit 42 says it evades standard binary scans and keeps those payloads out of reach of forensic analysts.
Why HTTP/3 makes the traffic harder to see
Both families use HTTP/3 as their main channel to their command servers. HTTP/3 runs on QUIC, which travels over UDP rather than TCP. Unit 42 notes that enterprise defenses cannot passively intercept that encrypted traffic without breaking the connection, and that some firewalls and intrusion detection engines still do not inspect UDP as closely as TCP, so outbound traffic on UDP port 443 frequently bypasses deep packet inspection.
A built-in safety switch and fake noise
Two PerfectPioneer variants contain a routine named isDangerousCommand that checks every incoming command against a hard-coded blocklist of destructive or forensics-heavy commands and refuses to run them, protecting the operators from their own mistakes. The list mixes Windows and Linux syntax, which Unit 42 says strongly suggests a parallel Linux version was in development. KUKURAT goes further with a camouflage engine that generates synthetic, benign-looking network, registry and file activity, burying real indicators of compromise in artificial noise.
Exchange servers and fake election documents
The main way in was exploiting unpatched Microsoft Exchange and web servers to plant web shells, including variants of ASPXSpy, Godzilla, Bysunue.aspx and Suo5. Phishing was a secondary route: victims downloaded executables disguised as PDFs from domains imitating official government websites, and one decoy impersonated a 2025 notice from Myanmar's Union Election Commission. In one case a fake installer ran a genuine Palo Alto Networks GlobalProtect setup in the foreground while dropping PerfectPioneer in the background.
Patch Exchange and look hard at outbound UDP 443
Since server exploitation was the primary entry point, internet-facing Exchange and web servers are the first thing to patch and to sweep for the web shells named above. The HTTP/3 channel is the bigger lesson: organizations whose inspection policies were written for TCP should check what their firewalls actually do with outbound QUIC traffic on UDP port 443. Unit 42's report carries the full technical breakdown and indicators.
A group that builds a safety switch to stop its own operators from tripping alarms is planning to stay a long time, and it is betting that defenders are still watching the wrong protocol.
This briefing is provided by IntelFusions for informational and defensive purposes only. It is based on sources assessed to be reliable at the time of writing, and analytic judgments carry the confidence levels indicated. Indicators of compromise are defanged; re-arm them only in controlled environments. IntelFusions is not affiliated with the organizations named and makes no warranty as to completeness or accuracy.