
Ransomware & Malware
https://hmmnm.com/living-off-trusted-infrastructure-c2-attacks-2026/
DragonForce tunneling ransomware C2 through Microsoft Teams TURN relays is the June 2026 proof that “living off trusted infrastructure” has gone from research concept to fielded attack. When malicious traffic looks like a Teams call, network detection dies.
Quick Answer
LOTI — Living Off Trusted Infrastructure — is the evolution of living-off-the-land: instead of PowerShell and WMI, attackers piggyback on services your firewall already trusts. In June 2026, Symantec revealed DragonForce’s Backdoor.Turn using anonymous Teams visitor tokens to route C2 through Microsoft’s TURN relay servers in a real attack on a major U.S. services company, after DLL sideloading (VirtualBox, DbgView) and BYOVD security-tool termination. Detections that still work: EDR on sideloading and driver-load anomalies, CASB/SaaS behavioral analytics on anonymous token generation and relay traffic, and zero trust verification that never whitelists a domain just because Microsoft owns it. Audit anonymous access on Teams/Slack now.
In June 2026, Symantec researchers revealed that the DragonForce ransomware group used a custom backdoor named Backdoor.Turn to tunnel command-and-control traffic through Microsoft Teams relay infrastructure — not a proof-of-concept, but a real-world attack against a major U.S. services company.
Living off trusted infrastructure (LOTI) is one of the most dangerous evolutions in modern attacks: rather than standing up their own flaggable infrastructure, threat actors piggyback on services your organization already trusts — Microsoft Teams, GitHub, Slack, Google Drive, even developer IDEs. Defenders see normal traffic from legitimate platforms while attackers exfiltrate data and deploy ransomware underneath.
The DragonForce attack chain was methodical:
The key innovation is step 4: TURN is the same standard protocol that enables calls when clients sit behind NATs — so all malicious traffic appeared as normal Teams activity to network defenders.
In 2025, Praetorian demonstrated “Ghost Calls” — hijacking temporary TURN credentials for Teams and Zoom to create stealthy communication tunnels through trusted conferencing infrastructure. It was a proof-of-concept then; DragonForce weaponized it.
LOTL abused built-in tools like PowerShell and WMI. LOTI expands the concept to entire cloud services and SaaS platforms — Teams and Slack for C2 tunneling, GitHub issues/commits/gists as covert exfil channels, Google Drive and OneDrive to blend stolen data with corporate traffic, Discord and Telegram APIs for C2 and distribution, and compromised AWS/Azure accounts to spin up temporary compute.
| Campaign | Trusted thing abused | Impact |
|---|---|---|
| DragonForce Backdoor.Turn | Teams TURN relays, anonymous visitor tokens | Ransomware C2 invisible to network defenses |
| FortiSandbox CVE-2026-39813 / -39808 / -25089 | The security product itself | Unauthenticated RCE in threat-detection pipelines |
| Malicious JetBrains plugins | IDE marketplace trust | 15+ plugins stole AI API keys from ~70,000 developers |
| Cisco Catalyst SD-WAN zero-day (CVE-2026-20245) | Network management plane | Root access to network infrastructure |
| ShinyHunters × Oracle PeopleSoft | Enterprise platform + KEV window | 100+ organizations claimed, incl. University of Nottingham |
The FortiSandbox case deserves emphasis — three critical flaws in a product whose job is detecting attacks, echoing the GreatXML/RoguePlanet Defender zero-days. The same week’s AUR hijack showed the package-ecosystem version of trusted-infrastructure abuse; see the full AUR analysis. And the AI-infrastructure variant — attackers weaponizing AI services themselves — is covered in Living Off the LLM.
The core problem is trust asymmetry. You pay for and configure these services to be accessible; your firewalls allow Microsoft 365, Google Workspace, and AWS; your SIEM whitelists their domains. Attackers routing through them inherit all that trust:
An attack approach where threat actors abuse legitimate, trusted services — Microsoft Teams, GitHub, Slack, Google Drive, cloud platforms — for command-and-control, exfiltration, and distribution instead of building their own infrastructure. It’s the evolution of living-off-the-land (PowerShell, WMI) scaled to entire SaaS platforms. Because firewalls and SIEMs whitelist these services, malicious traffic inherits their trust.
DragonForce’s Backdoor.Turn obtains anonymous Microsoft Teams visitor tokens and routes command-and-control traffic through Microsoft’s TURN relay servers — the same NAT-traversal infrastructure that carries legitimate Teams calls. To network defenders, the traffic is indistinguishable from normal Teams activity. Praetorian’s 2025 “Ghost Calls” research previewed the technique.
Bring Your Own Vulnerable Driver: attackers load legitimately signed drivers with known vulnerabilities — DragonForce used signed drivers from Huawei, Topaz Antifraud, K7 Security, and Tower of Fantasy — then exploit them to terminate security tools with kernel privileges. Countermeasures: Microsoft’s Vulnerable Driver Blocklist and driver-load anomaly monitoring.
Shift detection from network-level to identity and behavior-level: EDR for DLL sideloading and driver-load anomalies, CASB/SaaS behavioral analytics for anonymous token generation and unusual relay traffic, conditional access policies that eliminate anonymous sessions, and zero trust segmentation that inspects east-west traffic regardless of destination reputation.
{“@context”:”https://schema.org”,”@type”:”FAQPage”,”mainEntity”:[{“@type”:”Question”,”name”:”What is LOTI — living off trusted infrastructure?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”An attack approach abusing legitimate, trusted services — Microsoft Teams, GitHub, Slack, Google Drive, cloud platforms — for command-and-control, exfiltration, and distribution instead of attacker-owned infrastructure. It evolves living-off-the-land from built-in tools to entire SaaS platforms, letting malicious traffic inherit the trust of whitelisted services.”}},{“@type”:”Question”,”name”:”How does Backdoor.Turn use Microsoft Teams for C2?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”DragonForce’s Backdoor.Turn obtains anonymous Teams visitor tokens and routes C2 traffic through Microsoft’s TURN relay servers — the same NAT-traversal infrastructure as legitimate Teams calls — making malicious traffic indistinguishable from normal Teams activity to network defenders.”}},{“@type”:”Question”,”name”:”What is BYOVD and how did DragonForce use it?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”Bring Your Own Vulnerable Driver: attackers load legitimately signed drivers with known vulnerabilities — DragonForce used drivers from Huawei, Topaz Antifraud, K7 Security, and Tower of Fantasy — exploiting them to terminate security tools with kernel privileges. Counter with Microsoft’s Vulnerable Driver Blocklist and driver-load anomaly monitoring.”}},{“@type”:”Question”,”name”:”How do you detect attacks over trusted infrastructure?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”Shift from network-level to identity- and behavior-level detection: EDR for sideloading and driver anomalies, CASB/SaaS behavioral analytics for anonymous token generation and relay traffic, conditional access eliminating anonymous sessions, and zero trust segmentation inspecting east-west traffic regardless of destination reputation.”}}]}
Living-off-trusted-infrastructure C2 — attacker commands flowing through legitimate SaaS APIs, CDN fronts, and collaboration platforms — defeats IP-reputation and domain-blocking models by construction, because the infrastructure is exactly what enterprises allow. The detection doctrine that works: profile expected application behavior per process and per identity. The SaaS API that a compromised endpoint queries looks legitimate at the network layer; it looks wrong at the behavioral layer — the frequency, the API mix, the absence of corresponding user interface activity. Endpoint-to-cloud correlation (which process made which API call with which token) is the highest-value telemetry investment against this class.
The control stack layers beyond detection: egress design that routes SaaS traffic through authenticated proxies (identity-aware access rather than IP allow-lists), API-scoped service credentials so a stolen token reaches one function rather than a platform, and anomaly monitoring at the SaaS platforms themselves — the providers see cross-tenant abuse patterns that individual customers cannot. The SendGrid and Snowflake entries in this series document both halves of that stack failing and working respectively.
The strategic acknowledgment defenders owe themselves: trusted infrastructure is a shared resource, and its abuse is an externality of the trust model itself. The sector response — provider-side abuse detection, cross-industry threat intelligence on infrastructure-abuse patterns, and the behavioral telemetry doctrine above — manages the externality without abandoning the trust that legitimate operations require. The channel is not the vulnerability; the assumption that trusted channels carry only trusted actors is.
The closing synthesis for network defenders: trusted infrastructure abuse is not an argument against trust but a mandate for correlation — the day defenders instrument endpoint-to-cloud linkage as thoroughly as they once instrumented the perimeter is the day this C2 class loses its advantage. The telemetry exists, the correlation patterns are documented, and the adversaries are telling everyone exactly where to point the sensors by voting with their infrastructure choices.
Correlation is the counter to trust-abuse: instrument the linkage, watch the adversaries vote with their infrastructure, and the class loses its edge — telemetry first, advantage second.
Telemetry first, advantage second, and the day of correlation arriving one endpoint-cloud linkage at a time — the class loses its edge exactly as fast as the instrumentation gets built.