When the Admin’s Browser Becomes the Attack Path: Lessons from the TRUST#GRAFT WordPress Campaign | #hacking | #cybersecurity | #infosec | #comptia | #pentest | #ransomware


Defenders have spent years hardening the front door of WordPress sites with strong passwords, multi-factor authentication and login rate limiting. Despite this, a campaign we track as TRUST#GRAFT is exploiting a different weakness. Instead, they wait for a legitimate administrator to carry a valid session into the wrong web page. With WordPress powering roughly four in ten websites, techniques that work against its administrators can scale quickly.
TRUST#GRAFT is a browser-mediated attack, not a server exploit
The chain we analyzed begins with a malicious JavaScript loader running in a visitor’s browser. It overlays the page with a full-screen fake verification screen styled after a familiar “checking your browser” prompt. That screen hides the page underneath while the script performs additional activity in the background.
During that pause, the script fingerprints the environment, checking for automation tools, graphics and rendering characteristics, language, timezone and similar signals. It also solves a small proof-of-work challenge before the server returns to the next stage of the attack. This gating likely helps the operators filter out researchers and sandboxes and tailor what each visitor sees.
We also observed clipboard-based “ClickFix” behavior, where a lure persuades the user to paste an attacker-supplied command, along with attempts to harvest token-like values from browser storage. Recent research suggests the tactic is far from fringe. One vendor’s browser-layer telemetry shows ClickFix-style lures rising from roughly half of the browser-based attacks it detected in the second quarter of 2026 to about two-thirds by August.
The pivotal step comes next. When the loader runs on the affected WordPress site, it checks whether the visitor has an active administrator session with permission to install plugins. If so, it requests the plugin-upload page, extracts the legitimate security nonce WordPress issues, downloads a staged ZIP file and submits it through the normal upload endpoint. It then attempts activation through the standard link.
The point is easy to miss: this is not a WordPress authentication bypass. The chain depends on an administrator’s already trusted browser session. From the application’s perspective, the request looks like an admin installing a plugin, because in a technical sense that is exactly what’s happening.
 Persistence designed to outlast the original event
The installed plugin presents itself as a media optimization tool, but our analysis shows it is a backdoor. It injects remote JavaScript into public pages, retrieves its controller configuration with TLS certificate verification disabled, and exposes token-gated functions. One of those can set an authentication cookie for an administrator account, effectively restoring access without a password.
Two features make cleanup harder. First, the plugin hides itself from the standard plugin list, action links and update views. Second, it copies itself into the “must-use” plugins directory, which WordPress loads automatically and which cannot be disabled from the normal Plugins screen. An administrator who deletes the visible plugin may believe the problem is solved while a second copy keeps running.
What we still don’t know
Responsible analysis means being clear about its limits. We could confirm the loader’s behavior, the admin-session logic and the plugin’s capabilities. We could not confirm how the loader first reached victim pages. Compromised content, a malicious theme or plugin, a redirect chain and malvertising all remain plausible. The evidence we recovered also does not identify an operator, a country of origin or specific victim organizations, and infrastructure details alone shouldn’t be treated as attribution. Closing these gaps will take referrer data, server logs and page source from sites that actually served the loader, and we’d encourage affected site owners to preserve that evidence.
Practical steps for site owners and security teams
The defensive lessons apply well beyond this single campaign.

Separate privileged work from everyday browsing. Use a dedicated browser profile, or a separate browser, for WordPress administration. Avoid browsing unrelated sites while logged in to wp-admin, and log out when finished. If an administrator may have been exposed to suspicious content, invalidate their sessions.
Limit who can install code. Few accounts need permission to install plugins. Restrict that capability, consider requiring an approval step for new plugins and where your operations allow it, disable file modification through the dashboard entirely so that installs happen through a controlled deployment process.
Monitor the admin workflow as a sequence. A plugin-upload page request, an upload and an activation in close succession from one session is a strong behavioral signal, especially outside change windows. Individual events look routine, and the correlation is what exposes the abuse. File-integrity monitoring on the plugins and must-use plugins directories adds a second layer.
Audit for hidden persistence. Review the must-use plugins directory regularly and compare it against a known-good baseline. Check public pages for unfamiliar script tags, and review stored options and transients for unexpected remote URLs. Outbound connections from the web server to unfamiliar hosts deserve scrutiny.
Constrain what scripts can do. A well-tuned Content Security Policy and script allowlisting can limit the damage from injected JavaScript, though neither is a complete fix on its own.
Contain completely. If you suspect compromise, preserve evidence first, then remove every component, including any copy in the must-use directory. Rotate administrator credentials, invalidate all sessions and review user accounts for anything unexpected. Removing only the visible plugin invites reinfection.

The broader lesson
TRUST#GRAFT is a reminder that identity and session context are part of the attack surface. Multi-factor authentication protects the moment of login, but it does little once a session is established and a script in the same browser can ride on it. Treating administrator browsers as high-value assets, and administrative sessions as something to isolate and monitor, narrows the path attackers are counting on.
None of this requires exotic tooling. It requires discipline about who can install code, where privileged sessions live, and whether anyone would notice if a plugin appeared that nobody remembers approving.
_________
Author bio
Sina Chehreghani is Head of Threat Intel and Threat Research at Securonix, where he leads the company’s threat intelligence and research strategy. He focuses on helping security teams understand adversary behavior, emerging threats, and the practical detection and response implications of modern attack campaigns.
Sina brings deep experience across threat intelligence, threat research, adversary analysis, and security operations. His background gives him a practical view of how threat intelligence should inform detection engineering, investigation workflows, and executive understanding of risk.
At Securonix, Sina helps connect threat research to operational outcomes across the Unified Defense SIEM. His work supports better detection coverage, stronger threat context, faster investigations, and more actionable intelligence for SOC teams. He is especially valuable when a story needs a threat intelligence perspective that can bridge technical detail and security leadership relevance.
 
 
 

Join our LinkedIn group Information Security Community!

——————————————————-
Click Here For The Original Source.