By Hamza Faisal
Security researcher Orange Tsai has published the full details of a Microsoft Edge exploit chain that turns a malicious web page into code running outside the browser's sandbox on Windows — without using a single memory-corruption flaw. The chain, built solely from logic errors in how Edge and Windows handle ordinary features, won $175,000 at the Pwn2Own Berlin 2026 hacking contest in May.
Tsai, a researcher with Taiwan-based DEVCORE, released his technical write-up on October 28, five months after demonstrating the attack on stage. He says the work took two months and was completed manually, without AI assistance, and the finished exploit runs to fewer than 100 lines of code. Microsoft patched the affected Edge flaws the day after the contest, so an up-to-date browser is not at risk, and there is no evidence the exploit has been used against real users.
How a navigation flaw broke the website boundary
The chain begins in Edge's Conditional Access navigation handler, a component that helps enterprise users switch to the right browser profile when they visit Microsoft services. When the email address in a request matched the current profile, the handler started a new navigation to a destination taken from the request — but failed to bind that navigation to its original tab.
Chromium therefore used the active tab instead, letting JavaScript intended for one context run on whatever website the victim had open. That produced universal cross-site scripting, a flaw that breaks the boundary normally keeping websites apart and can bypass content security policies, since the navigation was started by the browser process itself. Tsai describes it as a new attack surface rather than a one-off bug, because it could reach sites an ordinary web flaw cannot touch.
Four more flaws removed the need for a click
Turning that primitive into a zero-click attack took four more weaknesses. Edge checked only the last committed URL of the destination window, allowing a trusted-origin bypass through a pop-up; a hard-coded allow-list entry let the attacker open a window without triggering the pop-up blocker; the OneAuth feature automatically signed users in on Microsoft sites; and a cross-origin resource sharing misconfiguration on a Microsoft asset domain exposed the signed-in account's email address, supplying the last piece of information the navigation trigger needed.
Cross-site script execution alone did not give operating-system code execution. The final stages of the chain worked around Chromium's blocks on privileged internal browser schemes and combined browser and Windows features to escape the sandbox entirely — the wall designed to keep a web page's code inside the browser and off the rest of the computer.
Why logic flaws worry browser makers
Browser escapes almost always rely on memory corruption, such as buffer overflows, which modern mitigations are built to catch. This chain used none, showing that weaknesses in how components interact can be just as dangerous as flaws in any single component. The research also earned Tsai the Epic Achievement award at the Pwnie Awards, alongside the Pwn2Own prize.
For users, the practical message is simple: the flaws are already fixed, so keeping Edge updated closes this path. For browser makers, the lesson is harder — ordinary features such as profile switching, automatic sign-in and pop-up handling can combine into a serious vulnerability when their boundaries are not checked at every step.



