As long as the attacker has at least one route into the organization at anyone time, the possibility of reinfection exists. So I guess it's wise to take a coordinated approach.
So, it means what you mean by 'externally' and how sophisticated you expect the malware to be.
After you get the infected machine, you pull out the drive and scan it externally, looking for bad hashes and files that shouldn't be there.
Yes, I can. Read up on modern digital forensics. Everything you do on a machine leaves a trace and there are ways to recover those traces and put together exactly what you did. That is exactly what Incident Response/Digital Forensics Firms do. An IR firm would never tell you to just reimage a machine when you're dealing with an advanced attacker. They'd want to go through, use the tools they have to identify exactly what happened on the machine and what other machines were compromised before they even started talking remediation. Wiping the one machine that you got an alert on would do absolutely nothing to solve the problem.
As a followup, both you and GP should read up on digital forensics from someplace OTHER than their marketing material...
> An IR firm would never tell you to just reimage a machine when you're dealing with an advanced attacker. They'd want to go through, use the tools they have to identify exactly what happened on the machine
Yes, I was talking more in the run-of-the-mill case sense, not regarding thorough forensic investigations. If you're actually going to investigate the incident deeper, you should at least get memory dumps, process dumps and an image of the machine and such. In my experience though, at most companies the SOP is just monitoring standard antivirus stuff and then when an infection comes up, it either gets automatically cleaned or someone goes over and fixes it with a manual scan or whatever. Which is completely inadequate.
Even working within the system, I'd say many attackers can remove traces such that many investigators won't find them, by doing things like deleting created logs, restoring file metadata to its original state, and writing over the erased evidence multiple times. (This perhaps assumes root access and a consumer-grade OS in default configuration.) It might lead to a suspicious state where the system has been running for hours with no artifacts that would routinely be left, but the investigator might not be able to determine much of what was done. It might be as simple as using a browser the investigators don't check: http://www.cbsnews.com/news/casey-anthony-detectives-overloo...
And the only way to clean up a compromised computer is a full reinstall. You can't possibly know what has happened on the compromised computer during the compromise. This is what the desktop support jockeys and most companies get wrong - obviously it's probably because of the cost associated with a full reinstall, but it doesn't make it any less valid. If it costs too much, companies should then focus on preventing machines getting compromised in the first place.
Think about it, if I ask you to hand over your laptop for say, an hour, during which I have completely free reign over it, can you tell me everything I've done during that hour and all the backdoors installed, if any?
And to nitpick, obviously these days not even a full reinstall might do it when there's BIOS viruses and even hard drive firmwares can be compromised etc. of course.