Rapid7 Reports North Korean Linux Espionage Toolkit Targeting Load Balancers
Dark Reading reports that a likely North Korean espionage group used a previously undocumented Linux toolkit to compromise HAProxy load balancers and access network traffic. The reported targets were South Korean media and automotive organizations; some intrusions may have been active since early 2025.
Rapid7 assessed the activity as linked to North Korean APT groups with medium confidence. Its assessment cited target selection, simple obfuscation, and command-and-control infrastructure associated with APT37, also known as InkySquid, ScarCruft, and Ricochet Chollima. The campaign highlights the security importance of network appliances: they can be consequential footholds, and their evidence may be overlooked when incident response focuses on servers and endpoints.
Read Dark Reading’s report and Rapid7’s analysis.
What the reporting describes
According to Dark Reading’s account of Rapid7’s research, attackers compromised HAProxy software on victims’ network appliances and installed a Linux toolkit Rapid7 called TED. Reported activity included harvesting credentials, redirecting selected users, conducting drive-by-download attacks, and modifying log files to conceal activity. Rapid7 described the intrusions as long-running.
Rapid7 identified possible intelligence interests in both sectors: information control and counterintelligence related to media organizations, and manufacturing intelligence related to automotive firms. These are assessments of potential motives, not confirmation that particular information was taken.
Attribution is also an assessment, not a definitive identification. Rapid7 rated its North Korean attribution medium confidence and cited multiple indicators, including infrastructure associated with APT37. Defenders can use the reported appliance compromise, toolkit, techniques, and targeting to guide their investigations without assuming that every similar intrusion has the same origin.
Why load balancers matter
Load balancers distribute requests among services and sit between users and backend systems. Because they handle traffic and can have privileged roles in production environments, compromising one may allow an attacker to tamper with routing, observe or manipulate traffic available to the device, or use it as a foothold into adjacent systems. The consequences depend on configuration, access controls, network design, and how traffic is protected.
The report’s account of access to incoming and outgoing traffic is a reason to investigate, not proof that every communication was readable. Encryption, where it terminates, application design, and device logging all affect what an intruder could see or alter. The reported credential harvesting and user redirection provide specific behaviors for organizations to check against their own evidence.
Treat load-balancing platforms as managed assets. Track which appliances run HAProxy, their versions and custom components, reachable management interfaces, and the people authorized to change configurations. Infrastructure can become a blind spot when teams do not treat it as software running on a computer with privileges and network access.
Investigating a suspected appliance compromise
Preserve evidence from the appliance and surrounding systems before rebuilding or making changes that could erase useful traces. Examine unexpected binaries or modifications to the HAProxy installation, configuration changes, unusual processes and network connections, altered or missing logs, new accounts or keys, and unexplained outbound traffic. Compare the device with a known-good image or documented baseline, and review management-plane access and administrative authentication records.
Since the reporting describes log modification, a clean local log is not conclusive on its own. Correlate appliance records with independent sources such as network flow data, centralized logging, identity-provider events, backend server logs, endpoint telemetry, and configuration-management history. Check for unusual routing changes, new destinations, unexpected authentication patterns, and activity by infrastructure accounts. Preserve originals and document collection methods so responders can distinguish incident evidence from changes made during the response.
Trace the traffic path and determine where TLS or other encryption terminates, which services the load balancer could reach, whether credentials handled by the appliance may have been exposed, and whether users were redirected. Review the device’s service identity and the trust relationships of adjacent systems. If credentials may have been harvested, plan rotations, prioritizing privileged and service accounts, and check for reuse. Account for availability when containing a production appliance: coordinate isolation, failover, and replacement with operations teams.
Linux appliance security and practical defenses
Manage the appliance as a Linux system rather than an untouchable black box. Restrict management access to dedicated administrative networks or approved jump hosts, require strong authentication, remove unused accounts and services, and patch software and dependencies through a tested change process. Limit service privileges, separate management and data-plane exposure, and restrict outbound connections to those the appliance needs.
Maintain an inventory of software and a known-good configuration, including custom plugins and locally compiled code. Use integrity monitoring or periodic comparisons with trusted images to detect unexpected changes. Forward logs to a protected, access-controlled destination, and monitor for gaps as well as suspicious events. Centralized records make local tampering harder to conceal and support correlation, though they cannot guarantee that records are complete.
Apply a risk-based hardening baseline to the Linux distribution and HAProxy deployment. Review SSH settings and key management, remove obsolete protocols and unnecessary packages, and make configuration changes reviewable and attributable. Test restoration procedures and maintain a replacement path for critical appliances so teams can rebuild a trusted device and validate its configuration under pressure.
These are established security practices, not evidence that any one control would have prevented this campaign. Tailor them to the deployment’s topology and availability requirements, and test changes before applying them in production.
What defenders should take away
Include network appliances in threat modeling, vulnerability management, logging, and incident exercises. Confirm ownership and visibility, establish integrity baselines, restrict management access, and practice investigation and failover before an incident occurs. A device that handles traffic can be strategically valuable to an espionage operator even when it is not normally treated as an endpoint.
The reported targeting also makes business context relevant. Media organizations can assess source protection and communications security; manufacturers can examine exposure of engineering and production environments. Both should map how an infrastructure compromise could reach identity systems, cloud services, and business applications. Investigate the reported behaviors against organizational evidence rather than assuming either that every communication was readable or that an appliance compromise is a routine software issue.
The available reporting describes a Linux espionage toolkit and does not identify AI or an AI agent as part of the operation. Broader concerns about AI-assisted attacks and the security of AI agents may inform separate planning, but they should not be attributed to this campaign without evidence. The practical lesson here is that trusted appliances, strong identity controls, independent telemetry, and practiced response are essential to resilient communications.
rapid7 reports north korean linux espionage toolkit targeting
north korean espionage campaign targeted south korean media
Network appliance security