Opera 7.51 for Windows and 7.50 for Linux does not properly prevent a frame in one domain from injecting content into a frame that belongs to another domain, which facilitates web site spoofing and other attacks, aka the frame injection vulnerability.
Argument injection vulnerability in Opera before 7.50 does not properly filter "-" characters that begin a hostname in a telnet URI, which allows remote attackers to insert options to the resulting command line and overwrite arbitrary files via (1) the "-f" option on Windows XP or (2) the "-n" option on Linux.
Opera allows remote attackers to bypass intended cookie access restrictions on a web application via "%2e%2e" (encoded dot dot) directory traversal sequences in a URL, which causes Opera to send the cookie outside the specified URL subsets, e.g. to a vulnerable application that runs on the same server as the target application.
Opera Web Browser 7.0 through 7.23 allows remote attackers to trick users into executing a malicious file by embedding a CLSID in the file name, which causes the malicious file to appear as a trusted file type, aka "File Download Extension Spoofing."
Heap-based buffer overflow in Opera 6.05 through 7.10 allows remote attackers to cause a denial of service (crash) and possibly execute arbitrary code via a filename with a long extension.
The PluginContext object of Opera 6.05 and 7.0 allows remote attackers to cause a denial of service (crash) via an HTTP request containing a long string that gets passed to the ShowDocument method.
Cross-site scripting (XSS) vulnerability in Opera 6.0 through 7.0 with automatic redirection disabled allows remote attackers to inject arbitrary web script or HTML via the HTTP Location header.
Opera, probably before 7.50, sends Referer headers containing https:// URLs in requests for http:// URLs, which allows remote attackers to obtain potentially sensitive information by reading Referer log data.