On Fri, 4 Jun 2004, Vern Paxson wrote:
Some people regularly rebuild their Windows computer a few times a year.
Including recovering from a trashed BIOS?
As you point out in the paper, the BIOS scrambler attack is the one with the most variation between platforms. It could have a relatively low success rate. On the other hand, board and even system replacement occurs pretty frequently. Even without a superworm, computer repair depots return alot of computers to the factory instead of trying to fix them locally, especially if the computer is new/recent/under warranty. The newest computers are most likely to have a fast replacement cycle from inventory. I agree the BIOS scrambler is a particular nasty form of attack, but the current state of computer repair expertise means its not that different from the problems created by current viruses. In 2000 Intel needed to recall over one million motherboards already shipped to end-users due to a defect. Instead of fixing the defective otherboards, Intel offered to replace them with new motherboards. Analysts estimated that recall cost $300 million to $400 million dollars including the labor to replace the motherboard. That's less than $400 per defective motherboard. Your paper estimates it would cost more than double to replace a scrambled BIOS. Although some people have a personal attachment to their computers, business PCs are very fungible. If the PC was more than a few years old, the business probably has depreciated it, and may just replace it sooner than planned with newer (faster, more productive?) model. If the business has more than one computer, again the BIOS scrambler has the most variables, some of the other computers may be different vintage. The business may just use one of the working computers instead. The damaged computers don't need to be replaced with the exact same make and model, in the short term even older models may be sufficient or do all those executives and other people need their computers 24x7? On a CPU hour basis, PC's have a very low utilization. SETI@HOME may have fewer free CPU cycles to borrow because any working "personal" computers will be shared by multiple people timesharing until their own replacement computers arrive. According to the banking industry, on average 5% of the cash machines in the country aren't working on any particular day. Its even higher during a holiday weekend. What's the economic impact of 5% failure rate of cash machines? If 10% of the cach machines had their BIOS scrambled, would the impact be doubled? Instead of an economic loss, do you create a Y2K effect of companies accelerating the replacement of equipment and hiring consultants to fix problems creating a mini-economic boom. Captalism seems to make more money treating illness than preventing it.