Executive Summary
In early 2026, Kaspersky researchers uncovered Operation ShadowCloud, an active APT intrusion campaign targeting government networks across Asia and Africa. Threat actors abused the legitimate MeshCentral remote-management agent as an initial foothold, then deployed DLL sideloading techniques to load two novel backdoors—GitDoor and GraphDoor—into memory. These implants tunnel command-and-control traffic through trusted cloud platforms including GitHub repositories and the Ably messaging service, effectively blending malicious communications with normal SaaS traffic and evading conventional network-based detection. The modular toolset supports keylogging, credential dumping via Mimikatz, screen capture, file exfiltration, proxy tunneling, and in-memory COFF shellcode execution, enabling deep and persistent access to sensitive government infrastructure.
Operation ShadowCloud represents a maturation of cloud-service-abusing C2 techniques previously observed in campaigns linked to Chinese-speaking threat groups. By routing implant traffic through GitHub APIs and Ably channels, attackers render perimeter-based detection largely ineffective, underscoring the urgent need for egress filtering, multicloud visibility, and zero trust segmentation in government and critical-sector networks.
Why This Matters Now
Nation-state actors are increasingly weaponizing legitimate cloud platforms—GitHub, SharePoint, Ably—as covert C2 channels, making traditional signature-based and perimeter detection obsolete. With attribution indicators pointing to Chinese-speaking APT groups and victims concentrated in government sectors across Asia and Africa, organizations must urgently prioritize cloud egress control, east-west traffic inspection, and zero trust architecture to counter this rapidly evolving threat vector.
Attack Path Analysis
Operation ShadowCloud began with attackers leveraging a pre-existing infection or unknown initial access vector to deploy the MeshCentral Agent onto Taiwanese government systems, enabling remote control and staging of DLL sideloading chains that loaded encrypted GitDoor and GraphDoor backdoors into memory. Once established, the threat actor escalated privileges using Mimikatz-based credential dumping, PetitPotam NTLM coercion, and token abuse techniques, then moved laterally across the network using proxied SOCKS5 tunnels via modified Erfrp and Ably channel proxies while deploying modular shellcodes for process injection into remote processes. Command and control was conducted covertly through GitHub repositories and the Ably messaging service, blending malicious traffic with legitimate cloud API calls to evade network detection. Sensitive data including credentials, keystrokes, browser cookies, screenshots, and files were exfiltrated via AES-256 encrypted payloads uploaded to private GitHub repositories and Ably channels, with secondary exfiltration to upload.gofile.io. The ultimate impact included persistent surveillance, credential harvesting, and covert long-term access to government networks across Asia and Africa.
Kill Chain Progression
This analysis maps confirmed threat intelligence to the full cloud kill chain to show where defensive gaps would emerge as an attack progresses.
Initial Compromise
Description
Attackers deployed the MeshCentral Agent (MD5: 74b025084026d8d9a64661bf995d835b) onto a Taiwanese government endpoint, likely by reusing a prior infection already present on the network. The agent connected to a hardcoded C2 at wss://27.102.66[.]162:443/agent.ashx, granting the threat actor full remote desktop, command-line, and file management access. Using this foothold, the attacker dropped DLL sideloading components disguised as legitimate binaries such as Sandboxie Start, Python, OpenJDK, and Sysinternals ZoomIt to load malicious loader DLLs (SbieDll.dll, python39.dll, jli.dll, WINMM.dll, libcef.dll) that decrypted and injected the core backdoor payloads into memory.
MITRE ATT&CK® Techniques
Phishing / Prior Infection Reuse (Initial Access via Trusted Agent)
Hijack Execution Flow: DLL Side-Loading
Process Injection: Dynamic-link Library Injection
Web Service: Bidirectional Communication (GitHub/SharePoint/Ably as C2)
Obfuscated Files or Information: Encrypted/Encoded File
Command and Scripting Interpreter: Windows Command Shell
OS Credential Dumping: LSASS Memory (Mimikatz / sekurlsa::logonpasswords)
Automated Collection (Keylogging, Screen Capture, Process Enumeration)
Exfiltration Over C2 Channel (AES-encrypted data upload to GitHub/Ably)
Create or Modify System Process: Windows Service
Scheduled Task/Job: Scheduled Task
System Binary Proxy Execution (Legitimate binaries used for DLL sideloading)
Adversary-in-the-Middle: LLMNR/NBT-NS Poisoning / NTLM Relay (PetitPotam EFSRPC abuse)
Reflective Code Loading (In-memory shellcode/COFF injection without disk writes)
Unsecured Credentials: Credentials In Files (Browser cookie theft via CookieKatz)
Potential Compliance Exposure
Mapping incident impact across multiple compliance frameworks.
NIST SP 800-53 Rev. 5 – Malware Protection, System Monitoring, Boundary Protection, Process Isolation
Control ID: SI-3, SI-4, SC-7, SC-39
CISA Zero Trust Maturity Model 2.0 – Network Segmentation and Application Allowlisting
Control ID: Network Pillar – Macro/Micro Segmentation; Application Workload Pillar – Authorized Software
NIS2 Directive (EU) 2022/2555 – Incident Handling, Supply Chain Security, and Network Monitoring
Control ID: Article 21 – Cybersecurity Risk Management Measures
ISO/IEC 27001:2022 – Vulnerability Management, Security Monitoring, and Web Content Controls
Control ID: A.8.8 Management of Technical Vulnerabilities; A.8.16 Monitoring Activities; A.8.23 Web Filtering
DORA (EU Regulation 2022/2554) – ICT Incident Detection Capability and Operational Resilience Testing
Control ID: Article 10 – ICT-related Incident Detection; Article 13 – ICT Business Continuity
PCI DSS 4.0 – Anti-Malware Controls and Audit Log Monitoring
Control ID: Requirement 5 – Protect All Systems Against Malware; Requirement 10 – Log and Monitor All Access
NYDFS 23 NYCRR Part 500 – Privileged Access Monitoring and Cybersecurity Incident Response
Control ID: Section 500.14 – Training and Monitoring; Section 500.16 – Incident Response Plan
Sector Implications
Industry-specific impact of the vulnerabilities, including operational, regulatory, and cloud security risks.
Government Administration
Operation ShadowCloud APT backdoors GitDoor and GraphDoor directly targeted Asian and African government networks, exploiting cloud C2 channels with lateral movement and exfiltration risks mapped to NIST.800-53 and ZTMM compliance frameworks.
Information Technology/IT
IT infrastructure faces critical APT backdoor exposure via DLL sideloading, in-memory shellcode injection, and cloud-blended C2 traffic evading east-west traffic controls and egress security policy enforcement across hybrid environments.
Computer/Network Security
Security operations confronting Operation ShadowCloud must address GitHub and Ably-tunneled C2 evasion, credential dumping via Mimikatz, and multicloud visibility gaps highlighted across ZTMM, PCI, and HIPAA compliance mappings.
Telecommunications
Telecom networks face elevated APT risk from encrypted C2 traffic blending with legitimate cloud services, MeshAgent remote access abuse, and proxy plug-ins undermining egress filtering and zero trust segmentation controls.
Sources
- Operation ShadowCloud: cloud service–based backdoors targeting government networks in Asia and Africa. Part 1https://securelist.com/operation-shadowcloud-gitdoor/122004/Verified
- MeshCentral - Open Source Remote Monitoring and Managementhttps://meshcentral.com/Verified
- CISA Known Exploited Vulnerabilities Cataloghttps://www.cisa.gov/known-exploited-vulnerabilities-catalogVerified
- Sophos Threat Research - APT Activity Brief (referenced in Securelist article)https://news.sophos.com/en-us/Verified
- MITRE ATT&CK - DLL Side-Loading (T1574.002)https://attack.mitre.org/techniques/T1574/002/Verified
- MITRE ATT&CK - Ingress Tool Transfer via Cloud Storage (T1105)https://attack.mitre.org/techniques/T1105/Verified
- NVD - National Vulnerability Database Searchhttps://nvd.nist.gov/vuln/searchVerified
Frequently Asked Questions
Cloud Native Security Fabric Mitigations and ControlsCNSF
Based on the attack progression modeled above, these are the defensive controls that would constrain each stage.
Aviatrix Zero Trust CNSF is directly relevant to Operation ShadowCloud because the attacker's progression depended entirely on unrestricted lateral movement, unmonitored outbound cloud API traffic, and open egress paths to GitHub and third-party file hosts. By enforcing workload-level segmentation and identity-aware egress controls, CNSF would likely reduce the attacker's reachability across internal segments and constrain the exfiltration channels used to maintain covert persistence.
Control: Cloud Native Security Fabric (CNSF)
Mitigation: The attacker's ability to establish an outbound WebSocket tunnel to a hardcoded external IP would likely be constrained by fabric-level visibility and policy enforcement, reducing the reliability of the initial C2 channel and limiting the foothold's ability to stage follow-on payloads.
Control: Zero Trust Segmentation
Mitigation: Zero Trust segmentation would likely limit the attacker's ability to relay coerced NTLM authentication to attacker-controlled destinations by restricting which workloads and network paths can reach external or lateral targets, reducing the effective scope of credential harvesting across the broader environment.
Control: East-West Traffic Security
Mitigation: East-west traffic enforcement would likely reduce the attacker's ability to pivot freely across internal network segments via injected SOCKS5 proxies, constraining the reachability of compromised hosts to only explicitly permitted peer workloads and limiting how far laterally the tunneled traffic could propagate.
Control: Multicloud Visibility & Control
Mitigation: Multicloud visibility and control could reduce the attacker's ability to operate undetected by surfacing anomalous outbound HTTPS traffic patterns to cloud API endpoints like api.github.com and rest.ably.io from workloads that have no legitimate business reason to reach those services, potentially constraining covert C2 channel reliability.
Control: Egress Security & Policy Enforcement
Mitigation: Egress policy enforcement would likely constrain the attacker's ability to exfiltrate data to multiple external destinations — including GitHub, Ably, and upload.gofile.io — by limiting outbound connections from internal workloads to only pre-approved external endpoints, reducing the number of viable exfiltration paths available.
Had prior CNSF controls constrained lateral movement and egress paths, the attacker's sustained access would likely have been limited to a smaller subset of workloads, reducing the volume of credentials, keystrokes, and files exposed and narrowing the scope of long-term surveillance achievable across the government network.
Impact at a Glance
Affected Business Functions
- Government Intelligence and Classified Communications
- National Security Operations and Diplomatic Networks
- Public Sector IT Infrastructure and Identity Management
- Interagency Data Sharing and Internal Government Portals
Estimated downtime: N/A
Estimated loss: N/A
The campaign specifically targets government sector organizations across Asia and Africa. Data at risk includes sensitive government credentials (usernames, NTLM hashes, cleartext passwords via Mimikatz-based credential dumping), active session tokens, keylogged communications from government officials, screenshots of classified or sensitive workstation activity, full filesystem contents accessible via the file manager module, browser cookies from Edge and Chrome (potentially including authenticated government portal sessions), WhatsApp database decryption keys, and network topology details including ARP tables, active TCP connections, and DNS configurations. The modular shellcode framework also enables persistent lateral movement, service creation, scheduled task manipulation, and NTLM relay attacks via PetitPotam, suggesting prolonged, stealthy access to sovereign government infrastructure with high espionage value.
Recommended Actions
Key Takeaways & Next Steps
- •
Deploy Egress Security & Policy Enforcement with FQDN filtering to block unauthorized outbound connections to attacker-controlled GitHub repositories (api.github.com used with malicious PATs), Ably messaging endpoints (rest.ably.io), and opportunistic exfiltration services (upload.gofile.io) — differentiating sanctioned cloud API usage from covert C2 channels via application-layer policy enforcement rather than IP blocking alone.
- •
Implement Multicloud Visibility & Control with deep traffic observability to detect anomalous API call patterns characteristic of GitDoor/GraphDoor C2 — specifically high-frequency GitHub API issue label toggling, repeated PUT/GET operations to atypical repository paths with victim-unique filename patterns, and Ably channel message bursts — establishing behavioral baselines that trigger alerts on deviations consistent with cloud-service-based C2 beaconing.
- •
Apply Zero Trust Segmentation with least-privilege identity-based policies to prevent lateral movement enabled by SOCKS5 proxy injection and Erfrp tunneling — enforcing workload-level microsegmentation so that compromised endpoints cannot establish arbitrary east-west connections to other internal hosts, and restricting service-to-service flows to explicitly authorized paths only.
- •
Enable East-West Traffic Security inspection on internal network flows to detect and block process injection-facilitated lateral movement (CreateRemoteThread, APC injection, KernelCallbackTable abuse) and proxy tunnel establishment — monitoring for anomalous inter-workload traffic patterns such as unexpected SOCKS5 negotiation, shared memory segment creation for proxy configuration (Local\pc<PID>), and unusual named pipe activity indicative of WinPTY reverse shell deployment.
- •
Activate Cloud Native Security Fabric (CNSF) inline enforcement with real-time inspection at cloud network ingress and egress points to intercept encrypted payload delivery disguised as legitimate cloud API traffic — correlating control-plane telemetry with data-plane anomalies such as Base64-encoded AES-encrypted content in GitHub commit messages, XOR-obfuscated filenames following victim-unique-code patterns, and Ably channel subscriptions from enterprise endpoints to external application IDs not associated with sanctioned business workflows.



