Browser

Offensive techniques against web code running in the browser: cross-site scripting and its execution contexts, cross-site request forgery, clickjacking and UI redress, and abusing or bypassing the same-origin policy, CORS, PostMessage, WebSockets, and the Content Security Policy that are supposed to contain untrusted web code.

1 phút đọccập nhật 6 thg 10, 2026florianamette1 người đóng góp
Hỏi trợ lý của bạn (MCP)Sửa trên GitHub

The browser runs untrusted code from every site at once and is supposed to keep each origin isolated from the others and from the user's data. Client-side web attacks live in the gaps of that model: getting attacker script to run in a trusted origin, making the victim's browser act against its will, and abusing the cross-origin messaging and resource-sharing mechanisms that relax isolation on purpose.

The surface#

  • Cross-site scripting: running attacker script in a trusted origin (reflected, stored, and DOM-based), its execution contexts and sinks, and the filter and sanitizer bypasses that reach them.
  • Request forgery and UI redress: cross-site request forgery, clickjacking, and other ways of making the victim's authenticated browser perform actions.
  • The origin model: the same-origin policy and its boundaries, CORS misconfiguration, PostMessage and cross-document messaging, WebSockets, and cross-site leaks.
  • Containment: bypassing or weakening the Content Security Policy and other client-side defenses, and abusing browser storage and extensions.

Seams#

Server-side injection, authentication, and logic flaws, including where a web application's server reflects or stores the data that becomes an XSS payload, are under Server > Web; this area is the browser-side execution and the origin model.

References#