This article is more than 1 year old
MS security patch fails on local files
Not an unmitigated disaster, but you be the judge
Updated The MS patch intended to fix a data binding flaw in IE, which enables a script to call executables on your Windows machine using the object tag, does not protect against malicious files launched from a local directory.
We had several anecdotal reports of trouble -- enough to cause alarm -- and went so far as to wipe a perfectly good Linux image off one of our machines and do a clean install of Win-XP. We then installed the relevant Microsoft Critical Security Update, and then ran an example script which harmlessly calls calc.exe against the machine.
We were able to run the progie without the slightest impediment from a local directory. If we hadn't tested the Critical Security Update with a live exploit, we'd never have known of its limitations.
However, with further tinkering we were unable to run the script from a remote location (i.e., a Web page) or in the body of an e-mail or as an e-mail attachment once the patch was applied.
Apparently, the patch approaches the data binding issue indirectly, by restricting script permissions in the Internet zone.
We questioned whether this approach offers adequate protection, or if it's really just a partial workaround.
"The patch prevents the exploit scenario as originally presented, i.e., visiting a malicious Web site which launches an exe on your system," MS Trustworthy Computing Team Senior Security Program Manager Eric Schultze told us.
"The patch prevents what users are concerned about: they shouldn't fear to install it thinking that it doesn't fix the problem. It in fact does fix the exploit scenario."
However, it seems reasonable to imagine that that quite a few people would copy an HTML attachment to a local directory and open it if their e-mail settings prevented them from seeing what they'd received. One might well say, 'well, it's only a little Web page, after all...where's the harm in that?'
"The patch does exactly what the bulletin makes it out to do, though whether this is the user's expectation is another matter," Schultze allows.
So I'd like to query our readers on this. E-mail me and let me know if you think the patch is adequate and we're making too much of too little here, or if you think it should be extended to the local computer zone, or even be re-worked completely to eliminate the data-binding/object-tag vulnerability altogether.
I'll post a sampling of your replies in a few days' time. ®