New research reveals that content within emails can bypass their message boundaries and interfere with webmail interfaces This article explores html emails sandboxed. . Across various attack chains involving Outlook, Gmail, Fastmail, Proton Mail, Yahoo Mail, and AOL Mail, these techniques enable capturing passwords, seizing third-party accounts, leaking tokens, hijacking trusted UI actions, and manipulating AI tools to read email.
The researcher noted that Fastmail resolved two CSS mutation bugs and a Proton Mail proxy bypass issue after retesting, while Outlook label-jacking and Gmail's image-set() bypass remained unaffected when the research was published on August 6. For webmail services, the research suggests placing HTML emails within sandboxed iFrames and strictly limiting CSS, custom attribute usage, select menu access, and image requests.
The investigation takes two routes: exploiting HTML and CSS features that are already permitted by webmail, or creating a gap where a sanitizer accepts something while the actual output differs. Allowed label elements enable external controls beyond messages, while JavaScript in applications turns sanitized custom attributes into new DOM nodes with CSS outside the sanitizer’s whitelist. The security measures begin with stringent isolation, followed by character-based input filtering for CSS validation, checks to prevent CSS gadgets before accepting custom attributes, blocking select menus and potentially harmful selectors, and preventing controlled image requests while allowing only those from a pre-approved domain list.
