Vishing crews target cloud data with account takeovers and extortion

src-5b893c16

Com-affiliated threat actors are using phone-based social engineering to steal access to corporate accounts, extract cloud data and pressure victims with extortion threats, according to incident response and threat research observations published by Unit 42. The campaigns have focused on data-rich organizations, particularly in financial services, healthcare and legal services, with most identified victims in the United States, the United Kingdom and Australia.

The activity has expanded since early 2026 and accelerated around August, researchers said. Rather than relying on file encryption, attackers have concentrated on accessing valuable business information and threatening to disclose it. Targets have included legal casework and strategy, financial and investment records, patient information, research, and large collections of customer and recruiting data.

Vishing leads to cloud account access

In a common entry method, an attacker impersonates internal IT support and tells an employee to resolve an authentication problem through a tailored web page. The page may act as an adversary-in-the-middle proxy, relaying credentials and multi-factor authentication approvals to the attacker in real time. Threat actors then use captured sessions to enter victim environments, often through residential proxy services or commercial VPNs.

Several activity clusters attributed to the broader Com ecosystem have been observed, including Bling Libra, also known as ShinyHunters, and clusters tracked as CL-CRI-1116 (Redact), CL-CRI-1147 (Pink), CL-CRI-1182 (Falcon), and CL-CRI-1192 (Helix). Researchers also identified two clusters in September: CL-CRI-1264 (Vortex) and CL-CRI-1274 (Cinder). The clusters share some infrastructure and techniques, but their operations are not identical.

Vortex activity has primarily targeted healthcare, pharmaceutical and other medical organizations. In observed incidents, attackers enrolled unauthorized MFA devices soon after gaining access, sometimes reusing a device named “Pixel 10a” across compromised accounts. They used Python-based automation to collect Microsoft 365 mailbox, OneDrive and SharePoint data. Afterward, they sent internal Microsoft Teams messages alleging data theft and directing employees to a Tor-hosted negotiation site.

Cinder operators also used vishing, but in observed cases they replayed a persistent “keep me signed in” session token to avoid another Duo MFA prompt. They then used scripted requests and the Microsoft Graph SDK to access and exfiltrate SharePoint and OneDrive content. Extortion messages were sent through victims’ email and Teams accounts.

Browser extension adds another theft channel

In an investigation associated with CL-CRI-1192, researchers found a malicious browser extension masquerading as “Google Chrome.” Versions 1.6.111 and 1.6.112 were identified; the extension was not distributed through the Chrome Web Store. The first observed sample was hosted on an attacker-controlled domain and required manual installation through the browser’s Developer Mode.

The extension could log keystrokes, collect browser storage and authentication cookies, scan pages for QR codes, and stream webcam feeds. Its remote-control functionality, implemented through Chrome DevTools Protocol, could also capture screenshots, interact with pages, access clipboard contents and navigate to websites. Researchers assessed with moderate confidence that CL-CRI-1192 developed the extension as an additional way to obtain credentials from Chromium-based browsers, including Chrome and Microsoft Edge. They said they had no evidence that other Com-affiliated groups were using it.

Defensive priorities for identity and data controls

Researchers observed attackers establishing persistence through unauthorized MFA devices, workstation objects, temporary cloud credentials, third-party OAuth applications and secondary accounts. They also monitored authentication and mailbox activity, changed calendar settings or mailbox rules, and in some cases cleared security alert events. Exfiltration methods included native SaaS features, REST APIs, scripted HTTP requests, Rclone and transfers between cloud services.

Organizations should treat help desk authentication and identity changes as high-risk control points. Unit 42 recommends strengthening employee verification procedures and adopting phishing-resistant MFA, including FIDO2-based authentication or passkeys. Security teams should also restrict browser extensions through allowlists, monitor for unexpected MFA-device enrollment and OAuth consent, and alert on unusual access or bulk exports from Microsoft 365 and other cloud collaboration platforms.

Monitoring should account for the attackers’ use of organization-specific authentication pages and short-lived subdomains, as well as proxy-based access that can make activity appear to originate outside a victim’s normal network. Repeated targeting is also a concern: researchers reported that some organizations faced renewed attempts weeks after an earlier operation failed. A response plan should therefore include rapid session revocation, review of identity persistence mechanisms, and assessment of mailbox, OneDrive and SharePoint activity after suspected account compromise.

The observed campaigns show how attackers can turn legitimate cloud access into a data-theft and extortion operation without deploying ransomware. For defenders, visibility into identity events, browser activity and data movement is essential to detecting an intrusion before stolen information becomes leverage against the organization.