A sophisticated and increasingly prevalent attack chain has been observed where threat actors abuse Microsoft Teams’ external collaboration capabilities to impersonate IT or helpdesk personnel, social-engineer users into granting remote desktop access, and then systematically pivot across the enterprise toward domain controllers, sensitive data stores, and cloud infrastructure. Unlike traditional email phishing, this campaign exploits a collaboration workflow that users instinctively trust, an inbound “support call” from what appears to be an internal IT team
Once the user grants remote access via Quick Assist or similar tools, attackers move fast: reconnaissance takes under two minutes, payload staging follows immediately, and lateral movement toward domain controllers can occur within the same session. The campaign concludes with Rclone-based exfiltration of business-relevant documents to attacker-controlled cloud storage
Any organisation using Microsoft Teams with external collaboration enabled is a potential target. The attack requires no vulnerability, no malware delivered via email, and no exploit, only a user who trusts an unsolicited support interaction
Technical
Details
Threat Overview
The intrusion chain begins with a cross-tenant Microsoft Teams message where the attacker impersonates helpdesk personnel using lures such as “Microsoft Security Update”, “Spam Filter Update”, or “Account Verification”. Voice phishing (vishing) is often layered on top to increase urgency and compliance. In some cases, this initial contact is preceded by a mailbombing event, followed by a “helpdesk assistance” message designed to overwhelm the user and increase the likelihood that they accept external support. Critically, Teams presents multiple security warnings at first contact, including external sender labels, Accept/Block prompts, spam/phishing indicators, and Safe Links checks, but this attack succeeds only when users actively bypass those controls
After consent is obtained, attackers establish interactive control via Quick Assist, perform rapid reconnaissance, stage DLL sideloading payloads, establish command and control infrastructure, move laterally via WinRM, deploy persistent RMM tooling, and exfiltrate data using Rclone, all within a single intrusion session
Technical Analysis
This intrusion chain begins with cross-tenant social engineering in Microsoft Teams, where external actors impersonate internal IT or helpdesk personnel to establish trust and prompt users into initiating remote support sessions. Once the user accepts the interaction and launches a tool such as Quick Assist, the attacker gains interactive control of the endpoint under the guise of legitimate support activity
After initial access, the attacker performs rapid system reconnaissance to validate privileges, domain context, and network reachability using native Windows utilities and lightweight command execution. This phase is typically short-lived and designed to quickly determine whether the host provides a viable pivot point within the enterprise environment
The attack then transitions into execution and staging, where trusted, signed binaries are abused for DLL side-loading to execute attacker-controlled payloads under legitimate process contexts. Payloads are often placed in writable directories such as ProgramData and executed through renamed or abused vendor applications to blend into normal software behavior
To maintain persistence and operational flexibility, attackers may store encrypted configuration data within registry keys and decrypt it at runtime, reducing reliance on disk-based artifacts. Command and control is established over HTTPS, often using infrastructure that mimics legitimate software update traffic to evade network-based detection
For lateral movement, adversaries leverage credential-backed remote administration protocols such as WinRM to pivot across domain-joined systems, with a focus on identity infrastructure and high-value assets. In parallel, additional remote management tools may be deployed to maintain redundant access paths and ensure persistence even if initial access vectors are disrupted
In the final stage, targeted data exfiltration is performed using synchronization utilities such as Rclone, with filters applied to selectively extract business-relevant data while minimizing detection footprint. Throughout the attack, the heavy reliance on legitimate tools and native protocols enables adversaries to blend into normal administrative and IT support activity, significantly complicating detection and response efforts
Attack Chain
#
Stage
1
Initial Contact via Teams
2
Remote Access Foothold
3
Reconnaissance
4
Payload Staging & DLL Sideloading
5
Loader State via Registry
6
Command & Control
7
Lateral Movement via WinRM
8
Persistent RMM Deployment
9
Data Exfiltration via Rclone
The stages are described separately below:
Initial Access via Teams Social Engineering
External tenant impersonates IT/helpdesk in Microsoft Teams
User receives chat/call request appearing legitimate
Attacker may reinforce trust via vishing (voice-based social engineering)
User is convinced to launch Quick Assist or similar remote support tool
Remote Access Establishment
Execution of QuickAssist.exe
Consent prompts handled via consent.exe
Attacker gains interactive desktop control
Rapid transition from chat to system-level execution
Reconnaissance & Validation
Execution of:
whoami, ipconfig, nltest, route print
Registry queries for OS/domain context
Network reachability checks for lateral movement planning
Payload Staging & Execution
Dropped payloads in writable paths (e.g., ProgramData)
DLL side-loading via signed binaries:
AcroServicesUpdater2_x64.exe → msi.dll
ADNotificationManager.exe → vcruntime140_1.dll
DlpUserAgent.exe → mpclient.dll
werfault.exe → Faultrep.dll
Execution under trusted application contexts
Persistence & Loader Behavior
Registry-based encrypted payload storage (HKCU)
In-memory decryption of configuration data
Fileless execution patterns aligned with loader frameworks
Command & Control (C2)
Outbound HTTPS traffic over port 443
Cloud-hosted, dynamically rotating infrastructure
Beaconing disguised as legitimate update traffic
Lateral Movement
WinRM-based execution (TCP 5985)
Credential-backed movement across domain-joined systems
Targeting of identity infrastructure (e.g., domain controllers)
Data Exfiltration
Use of Rclone for selective synchronization
File-type exclusions to reduce detection footprint
Upload to external cloud storage platforms
MITRE ATT&CK Mapping
Tactic
Technique
Observed Behaviour
Initial Access
T1566.002
Phishing via spearphishing link: Teams message delivers social engineering lure
Execution
T1204.004
Malicious Copy and Pasteuser pastes clipboard content into Quick Assist or browser
Command & Control
T1219
Remote Access Software: Quick Assist, AnyDesk, TeamViewer used for initial foothold
Defense Evasion
T1574.001
DLL Sideloading: malicious DLLs loaded by signed legitimate executables
Defense Evasion
T1112
Modify Registry: encrypted C2 config stored in HKCU registry keys
Persistence
T1053.005
Scheduled Task: used for payload persistence alongside RMM deployment
Lateral Movement
T1021.006
Remote Services: WinRM: credential-backed pivot to domain controllers
Lateral Movement
T1072
Software Deployment Tools: Level RMM deployed via msiexec across hosts
Execution
T1059.001
PowerShell: used for download and staging activity
Discovery
T1082
System Information Discovery: whoami, systeminfo, hostname, nltest
Discovery
T1016
System Network Config Discovery: ipconfig, arp, route print
Credential Access
T1552.001
Credentials in Files: harvested to enable WinRM lateral movement
Exfiltration
T1567
Exfiltration to Cloud Storage: Rclone transfers to external cloud destination
This technique reflects a broader shift from traditional phishing to interactive social engineering attacks inside collaboration platforms. Rather than delivering payloads directly, attackers rely on users to initiate compromise themselves through trusted workflows
Key trends include:
Increased abuse of Microsoft Teams external collaboration
Heavy reliance on legitimate remote support tools
Growth in “helpdesk impersonation” as a primary intrusion vector
Blending of human manipulation with legitimate administrative tooling
Cloud storage services used for exfiltration (via Rclone)
Why This Is
Important
This campaign represents a meaningful evolution in enterprise social engineering: it requires no vulnerability, no malicious email attachment, and no exploit kit. The attacker’s primary weapon is the enterprise’s own collaboration infrastructure, and the “mechanism of delivery” is a user who believes they are receiving legitimate IT support
The business and operational risks are severe and compounding:
Domain controller compromise: WinRM-based lateral movement targeting identity infrastructure can result in full Active Directory takeover, enabling attackers to create persistent admin accounts, modify GPOs, and establish long-term footholds that survive endpoint remediation.
Credential exfiltration at scale: Harvested credentials from a single endpoint can provide access to dozens of downstream systems, VPNs, cloud workloads, and SaaS platforms.
Business data exfiltration: Rclone’s targeted file-type filtering indicates deliberate selection of commercially sensitive documents, contracts, financial data, and strategic plans, creating regulatory breach notification risk under GDPR and sector-specific frameworks
Stealth and dwell time: The use of signed legitimate executables (Quick Assist, WinRAR, Adersoft tools), blockchain-based C2 configuration, and commercial RMM platforms means this campaign consistently evades signature-based and network-level detection. Dwell time before detection significantly amplifies impact
Compounding threat actor handoffs: Microsoft has observed that in some campaigns, initial access brokers may hand off to different threat actors mid-intrusion, meaning later stages can differ materially in objective and tooling
This technique profile underscores how attackers increasingly weaponise the trust users place in enterprise collaboration platforms, and why identity-aware, behaviour-based detection across Teams, endpoint, and network telemetry is now a fundamental detection requirement, not an enhancement
How Wizard Cyber Can Help
24/7 monitoring through Microsoft Sentinel for cross-tenant impersonation patterns
Yara specialises in proactive threat hunting, security awareness, and cyber security education. She combines technical threat analysis with user-focused security initiatives to help organisations strengthen their overall cyber resilience. She holds Microsoft SC-200, AZ-500, and SC-300 certifications
Certifications: SC-200, AZ-500, SC-300
Threat Hunting Team
The Threat Hunting Team at Wizard Cyber is focused on proactively seeking out advanced threats that evade traditional security measures. Leveraging advanced analytics and deep knowledge of threat actor behavior, they uncover hidden risks within our clients' environments. This team's continuous monitoring and analysis ensure that any potential compromises are detected and neutralized before they escalate.