Japan's National Police Agency (NPA) and National Center of Incident Readiness and Strategy for Cybersecurity (NCSC) accused a China-linked threat actor named MirrorFace of orchestrating a persistent attack campaign targeting organizations, businesses, and individuals in the country since 2019.
The primary objective of the attack campaign is to steal information related to Japan's national security and advanced technology, the agencies said.
MirrorFace, also tracked as Earth Kasha, is assessed to be a sub-group within APT10. It has a track record of systematically striking Japanese entities, often leveraging tools like ANEL, LODEINFO, and NOOPDOOR (aka HiddenFace).
Last month, Trend Micro revealed details of a spear-phishing campaign that targeted individuals and organizations in Japan with an aim to deliver ANEL and NOOPDOOR. Other campaigns observed in recent years have also been directed against Taiwan and India.
According to NPA and NCSC, attacks mounted by MirrorFace have been broadly categorized into three major campaigns -
- Campaign A (From December 2019 to July 2023), targeting think tanks, governments, politicians, and media organizations using spear-phishing emails to deliver LODEINFO, NOOPDOOR, and LilimRAT (a custom version of the open-source Lilith RAT)
- Campaign B (From February to October 2023), targeting semiconductor, manufacturing, communications, academic, and aerospace sectors by exploiting known vulnerabilities in internet-facing Array Networks, Citrix, and Fortinet devices to breach networks to deliver Cobalt Strike Beacon, LODEINFO, and NOOPDOOR
- Campaign C (From June 2024), targeting academia, think tanks, politicians, and media organizations using spear-phishing emails to deliver ANEL (aka UPPERCUT)
The attacks are also characterized by the use of Visual Studio Code remote tunnels to establish covert connections, thereby allowing the threat actors to bypass network defenses and remotely control compromised systems.
The agencies also noted that they observed instances where the attackers stealthily executed the malicious payloads stored on the host computer within the Windows Sandbox and have communicated with a command-and-control server since at least June 2023.
"This method allows malware to be executed without being monitored by antivirus software or EDR on the host computer, and when the host computer is shut down or restarted, traces in the Windows Sandbox are erased, so evidence is not left behind," the NPA and NCSC said.
Update
In a follow-up analysis published on March 12, 2025, the ITOCHU Cyber & Intelligence team detailed the Windows Sandbox technique abused by MirrorFace in attacks targeting Japan.
"By default, Windows Sandbox is disabled. Therefore, after compromising a target machine, the attacker enables the Windows Sandbox feature. Since Windows Sandbox only becomes active after the host machine is restarted, the attacker creates a WSB (Windows Sandbox configuration) file on the system and then reboots the host machine," the company said.
"After the reboot, Windows Sandbox is available, and the malware (in this case, LilimRAT) is executed within the sandbox according to the WSB file's configuration, establishing communication with the C2 server."
ITOCHU also noted that proactive measures such as enhanced monitoring is of paramount importance since antivirus solutions and endpoint security software on the host machine may not detect threats within the sandbox.
(The story was updated after publication on March 13, 2025, to include additional analysis of the Windows Sandbox technique shared by ITOCHU.)