Re: Can a computer virus kill the CPU?

For a longest time, satellites, et al did not have computer
(microprocessors) or even encryption. Satellites were a gentleman's
game even during the Cold War. Most processing was performed on earth.
A satellite had only enough intelligence to maintain itself for short
periods. Electronics was the simplest and the least complex possible.

Later, simple processors were installed only for station keeping -
only to increase reliability. I suspect TDRSS (NASA's communication
satellite system) may have been a first to use intelligent processing
for data - satellite operating as a switch and not as a repeater.
However that is the trend. Electronics in aerospace is the simplest,
old, and well understood because reliability and being 'fully
qualified' are so critical. To put something more advanced means there
was no alternative. Only then was massive money spent to qualify a new
part.

When more processing power is required, an expensive process would be
used to qualify a newer and yet well proven processor. So designers
kept most intelligent functions on earth. Yes, slowly more robust
processors make it into space. But rarely is anything 'state of the
art' in aerospace. 'Reliability' is a far larger consideration which
often means old and well proven hardware. Processing power of one
astronaut's laptop is far more powerful than all Shuttle computers
combined. That laptop need not be reliable. Those shuttle computers
must be extremely reliable - and still they sometimes fail. Therefore
the Shuttle has five computers making flight decisions. A poll
determines which three will be believed. Again, it's about
reliability and why those computers have so little processing power.


Viruses that attack firmware have even been shipped (unknowingly) by
the computer manufacturer. Such viruses, although rare, have been
observed (and manufacturers don't like to talk about it). But again,
no hardware damage.

Cited previously was the case where a monitor signal too high in
frequency could cause monitor damage. So yes, sometimes a designer
does make a mistake, the lessons learned, and hardware fixed so that
software cannot cause damage.

We also built customized equipment where a software bug could cause
relay damage. But again, this was customized equipment AND restricted
users also had no access to embedded software. Software was embedded
so that damage could not result from user programming. Even customized
equipment makes hardware as resilient as possible to software failure
or malware.

The context is not such custom equipment. This context is consumer
computers. Consumer electronics is designed so that software cannot
harm hardware. Consumer equipment must be that 'idiot proof'.

More complex and custom equipment can be harmed if the attacker has
very specific engineering knowledge of that item; a specific item
targeted. Specific engineering knowledge is what a virus writer must
have to specifically attack firmware for a specific machine. These
customized machines were designed so that software could not cause
hardware damage. But then some functions might not be so robust
because costs would be so high and because of the unique customized
nature of its function. It is possible to put a virus into a robot so
that robot destroys its own arms. But then good luck trying to do it.
It is possible, but ....

Supercomputers are a different environment. The days of customize
hardware (ie Cray XMPs) are gone. Same processor chips built for
commercial purposes now are used in supercomputers where numbers of
processors - not specialized functions - are the objective. Most
'souped up' chips are sold in the market. An example: Sony Playstation
3. This new hardware is also challenging how software can utilize the
power - creates a massive challenge to software engineers.

Don't assume that because it is a nuclear missile, then it must have
the most advanced electronics. Instead, it will have the most
'reliable' electronics. Each market has different parameter
requirements. Sony Playstation 3 would be far too unreliable for
aerospace use.

But consumer electronics is carefully designed so that software will
not damage hardware. Computers are legacy designs where most of the
machine is simply a faster version of the same thing done a decade ago.
That legacy requirement is important because of what the market
demands of consumer electronics - where everything is a black box and
the retail 'computer expert' does not even know what a power supply
does. In such markets, hardware is designed so that software (viruses,
bugs, and mistakes) cannot damage hardware.

mike4ty4@xxxxxxxxx wrote:
So then, basically, it would be very difficult to write a virus that
could attack 95% of the firmware out there and thus require
95% of infected systems to be sent back to the factory (which
is what I had in mind when I asked the original question -- a
virus that would demand factory return.), or have parts
replaced, etc., and almost impossible to damage hardware
physically with the virus, right?

You mentioned about spacecraft not using advanced Pentium
processors, only 8086s. But what if they needed more
computing power? 8086 is a very slow processor and is only
useful for relatively simple things. Even though they might not
need it now, what if they do in the future?

I also needed to know about the motivation to make the virus
thing, where I asked about the military. Could it be possible
that the military has a virus that can damage hardware or
otherwise mandate physical repair procedures?

Also, though, for military computers used for doing complex
calculations or other intense activities, including
supercomputers, is it possible that the chip companies
manufacture "souped-up" chips for THOSE computers?

.