@_jensec Shameless bit of self-promotion here: I've been working on https://t.co/kX7EDdIrdd for a few years now, and we're doing really solid, unique, interesting work to address challenges in the CVE ecosystem
I don't think this is as much validation as it is routing. A reverse proxy will likely use the initial host header to determine what the destination should be which will persist for as long as the connection is alive. However, I only see the SSRF occur in very specific scenario's: e.g. an upstream server takes arbitrary host headers but isn't bound to a known list, unlike the first proxy.
@slonser_@terjanq@serverinspector Ah I see, it's always been so obvious to me that cross-origin navigations are allowed that I never thought to check. Doesn't look like any kind of cross-origin interactions are allowed either
@terjanq@slonser_@serverinspector Under the assumption that the vulnerable page is frameable, an attacker could extract the blob URL via the technique in the original post, poll the attacker server (client-side) for the extracted blob URL, then navigate the top frame to the extracted URL
@slonser_@serverinspector If the vulnerable page is frameable, you can poll your server (or websockets), let the browser wait for the received blob, and redirect to it. Becomes more of a dev issue at that point
@fwrnr@yaswanth__03@Bugcrowd I should probably clarify that it doesn't "lead to open redirect". Assuming that the above applies here, then it's just incorrect usage of "open redirect" and a better title would be something along the lines of: "XSS allows page navigation"
Which is still an unusual PoC for XSS