Microsoft Edge Zero-Click Exploit Chains Five Flaws for Sandbox Escape
A zero-click exploit chain for Microsoft Edge has been detailed, chaining five vulnerabilities and two Windows features to achieve code execution outside the browser sandbox.
Security researcher Orange Tsai has revealed a sophisticated exploit chain for Microsoft Edge that bypasses the browser's sandbox to achieve code execution on Windows without relying on memory corruption vulnerabilities. The exploit, developed for Pwn2Own Berlin 2026, chains together five distinct vulnerabilities within the browser and leverages two Windows features to achieve its objective.
Tsai's research highlights how logic flaws in the browser's navigation handling, coupled with Content Security Policy (CSP) bypasses and misconfigurations in authentication and resource sharing, can be combined to enable universal cross-site scripting (UXSS). This allows malicious JavaScript intended for one website to execute in the context of another, breaking down the isolation that normally separates web content.
The exploit chain begins by targeting Edge's Conditional Access navigation handler. A flaw in how this handler manages tab context allows JavaScript to be injected into an active tab, even if it originates from a different, attacker-controlled context. This initial step is crucial for establishing the UXSS condition.
Further complicating defenses, the exploit bypasses Content Security Policy protections by exploiting a cross-window flow and a hardcoded popup exception. These mechanisms allow the attacker to circumvent checks that normally restrict navigation from untrusted sources or prevent unexpected pop-up windows, thereby enabling the execution of malicious code.
The chain also requires a signed-in Edge profile and access to its associated email address. The research demonstrates how Microsoft's OneAuth feature can automatically create and sign in a profile, and a cross-origin resource sharing misconfiguration on assets.msn.com exposes the signed-in email to the attacker's website. This combination removes the need for any user interaction after the initial loading of malicious content, earning the exploit its "zero-click" designation.
Once cross-site scripting is achieved, the exploit targets Edge's Immersive Reader feature, specifically its read:// address scheme. By exploiting path traversal and null-byte handling, the attacker can write files outside the intended temporary directory and remove file extension restrictions. This capability is then used to modify Edge's protocols.json configuration file.
Finally, by altering protocols.json, the attacker can force the browser to invoke a Windows handler, such as the telnet client. By placing a small executable in a directory that Windows searches for executables, the attacker achieves code execution outside of Edge's sandbox, demonstrating a powerful technique that relies on logic flaws rather than memory corruption.
This research underscores the importance of holistic security testing, emphasizing how the interaction between browser features, identity management, and operating system integrations can create exploitable attack surfaces. Users are advised to keep their Microsoft Edge browsers updated to patch these vulnerabilities and to understand that browser isolation is not an absolute defense.