Dale Peterson, a long time reader and cybersecurity
blogger/expert, left
an interesting comment on yesterday’s post about industrial
feedlot vulnerabilities. He noted that:
“A lot of control systems may not
be critical infrastructure but have a big impact on an individual or business
if compromised.
“A few years back we did an assessment at a prominent University. One of their big concerns was a multi-hour electrical outage or HVAC failure could wipe out numerous grad and doctoral students' research projects.”
“A few years back we did an assessment at a prominent University. One of their big concerns was a multi-hour electrical outage or HVAC failure could wipe out numerous grad and doctoral students' research projects.”
Dale is absolutely correct; everyone that owns any kind of
control system has something to lose if that control system is compromised,
even if it is nothing more than the inconvenience of not being able to open
your garage door. Of course, the same could be said about concerns about the
general reliability of the system.
Risk Assessment
This is the reason that all control system owners, down to
the garage-door opener owners, need to conduct a risk assessment for their
systems. I think that a realistic risk appraisal by most ICS system owners
would not show a high threat of terrorist attack. Critical infrastructure facilities
would probably be an exception and some other facilities where a specific group
would have an ax to grind with facility owner/operator (our feedlot example for
instance). On the whole, however, most facility owners do not have to worry
about terrorist cyber-attacks.
Two other types of outsider cyber-attacks should be
considered in any reasonable risk assessment; electronic thieves and ruthless
competitors. Electronic thieves may be after anything of value including ‘protection
payments’ for not shutting down the control system. Ruthless competitors (and
that includes some nation-states) could be after process information or be
trying to compromise the integrity of the control system to put competitive
production at a disadvantage.
There is one other form of outsider attacks that is probably
going to become more prevalent now that the vulnerabilities of control systems
and their internet accessibility are becoming better known; script kiddies.
These are frequently adolescent (not necessarily age defined) individuals
seeing what they can accomplish to make a name for themselves. As more ICS
attack tools become generally available on the Internet, the number and
exploits of these denizens of the dark side of the Internet will become more of
a problem for control system owners.
The most common form of cyber-attack for most facilities is
not an outsider. Most ‘attacks’ will come from within the firewall and may be
deliberate attacks by employees or contractors with personal grudges or, probably
more common, accidental ‘attacks’ where employees or contractors inadvertently do
something that has some sort of disruptive effect on the system. The last
category is probably the most common form of control system incident and needs
to be better documented.
Control System Vulnerabilities
All of the control system vulnerabilities that are reported
by folks like ICS-CERT, vendors (like the Siemens-CERT) and independent
security researchers (white hats) make any of the above described attacks
easier. As these vulnerabilities are discovered and mitigated (or mitigations are
developed) it is the responsibility of the ICS owner to ensure that the
mitigations and protective tools are applied to their systems.
Unfortunately, I would suspect that the vast number of
control systems do not have systems engineers available to track
vulnerabilities and implement protective mitigations. Large company systems
probably have some level of protective services available, but most small
company owners that employ the lower cost systems have no idea that the
vulnerabilities exist, much less how to protect them.
The Solutions Are Not Easy
The ‘easy answer’ would be to require vendors to push
vulnerability report and mitigation measures to the owners. There are a couple
of problems with this. First, many systems are not sold directly by the vendor
so they have no way to contact all of the owners. Systems where a direct push
of new versions and updates to the ICS (and we have seen more reports of this
type action lately on ICS-CERT advisories) is possible, the vendor runs the
risk of disrupting the actual operation of the control system.
Finally, the long time over which an ICS is used ensures
that there will be a turnover of knowledgeable employees on site and maybe even
of the management team while the system still runs. There are some unknown
number of systems where the current owners are just letting the system run,
hoping that nothing breaks down that their routine maintenance can’t address.
The long term solution is to engineer industrial control
systems with security part of the integral design. Even that won’t be a perfect
solution. It just takes too long for control systems to die. That and the fact
that even with security part of the design process, there will still be hole to
find and exploit. Just look at how long Microsoft has been working at their
security processes; they have their security updates down to just a couple of
times per month…
Moving Forward
No, everyone in the ICS sector needs to be more aware of the
security problems and there has to be better communications between everyone in
the community. Vendors need to reach out to owners. Owners need to network to
gain access to the necessary information. White Hats need to keep plugging away
at problem identification. And people like me need to keep bugging the world
about the problem.
1 comment:
Patrick - The flip side is the risk assessment might show little impact of a compromise. There are a huge number of control systems that fall into this category.
Consider a SCADA system that reads natural gas meters. If the meter has no control capability and is simply connected to a sensor or sensors taking measurements, the impact of a compromise would be one or many meter readings being incorrect.
The first reaction could be ... that could cost us a lot of money. However most of these systems have many other controls beyond the meter read via SCADA.
- they may have a physical read (a person goes out and looks at the meter) quarterly. The impact would be a loss of time value of money for those two months
- most of these billing systems have automated reasonableness checks. If the bill varies beyond a certain percentage it triggers an action, such as a manual check of the meter
- if you look at a lot of utilities terms of service you will see they have the right to extrapolate or interpolate usage if the SCADA read is not available on the billing day.
You are right on in focusing on risk. Nothing turns an Operations Team off more than a "security guru" coming in and emphasizing a vulnerability that has little impact.
Post a Comment