Back Infosecurity-Magazine Suspected Chinese APT Rebuilds EggStreme Malware After Exposure
A suspected Chinese advanced persistent threat (APT) group rebuilt parts of the EggStreme malware framework after it was exposed by Bitdefender. The company said the timing suggested its report was a likely trigger.
Bitdefender detailed the changes in an update to its cyberespionage research, first published in September 2025. It made no new attribution claim. It assessed with high confidence that the operation targeted multiple Philippine government organizations and with low to medium confidence that it targeted Taiwan, calling the latter a lead, not an established fact.
In the original research in September 2025, Bitdefender found a Chinese APT group had compromised a Philippines-based military firm using the EggStreme fileless malware framework.
What the Operators Changed
EggStremeFuel, one of the framework's components, changed how it protected its configuration in a build compiled 12 days after the original report, Bitdefender said.
Two weeks after that, the operators compiled a loader with a new cipher and payload format. That cipher was replaced again by late October 2025, swaps Bitdefender said offered little real protection but still defeated detections built for the old cipher.
The EggStremeAgent backdoor dropped its protocol for raw TCP connections still secured with mutual TLS, in which both sides present certificates.
Newer builds added Windows domain mapping and in-memory PowerShell, and none of those analyzed carried the keylogger. Two of them ran in passive mode, waiting for the operators to connect rather than calling out. Bitdefender first saw these builds after the report, though some were compiled earlier.
Bitdefender said the framework's design, which let operators swap capabilities with a new build, appeared to explain the speed.
What Defenders Can Still Track
Older EggStreme servers traced back to one certificate authority, created in 2019 and linked to 32 domains. No new domains appeared under it after the report. Newer activity suggested a shift to campaign-specific ones, which Bitdefender said was unconfirmed, though some links between old and new infrastructure remained.
Despite the rebuild, four behaviors stayed constant: DLL sideloading (getting a legitimate program to load malicious code), movement through file shares, in-memory execution and mutual TLS. Bitdefender said the operators "spent at the tools level and kept their methods."
Bitdefender's recommendations included restricting built-in Windows tools an environment did not need, detecting DLL sideloading by library name and location, finding unmanaged machines that could still reach workstations and watching for new inbound listeners on workstations.
The company argued that exposure could cost operators more than one campaign, since tooling is often shared. It said that raised a question handing such operations to autonomous AI tools, where the chance of discovery was high, though its data did not test it.
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
