Page 265 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

6. Cybersecurity 257

6. CYBERSECURITY
As mentioned in the previous sections, CPS and IoT technologies will make our life
easier and richer in opportunities. However, at the same time, such technologies are
prone to cyberattacks designed to steal precious information and harm people,
companies, and governments. In late September 2016 bot nets infected by the
infamous IoT malware Mirai [5] caused a severe Distributed Denial of Service
Attack (DDoS) that prevented major sites to correctly provide services to clients.
Cyberattacks using IoT devices are expected to steadily increase by taking
advantage of network scalability and vulnerability of IoT devices, an aspect that
is far from being fixed. Moreover, malware Mirai that was originally designed to
infect Linux-based IoT devices is still evolving, with its offspring targeting
Windows computers and possibly, in the future, also Android smartphones.
Cyberattacks to CPSs are of deeper concern once applications target our daily life.
Consider, for instance, an elderly care smart home with a sole senior person whose
behavior is continuously monitored for safety purposes. The monitoring system
consists of some surveillance cameras and various IoT sensors such as door lock
and proximity sensors deployed in rooms and corridors. Once the monitoring system
detects an emergency situation, an alert is issued and sent to caregivers and doctors.
Following a malware infection, the system control is completely taken by the attacker
who might decide to activate, for example, a DDoS attack, thus leaving the person’s
safety issue unattended. Needless to say, CPS designed to protect the person will, in
the best case, not work properly. It comes per se that smart cybersecurity systems
need to be considered in order to prevent such situations.

6.1 HOW CAN CPS AND IoT BE PROTECTED FROM CYBERATTACKS?

In order to effectively protect a given CPS or IoT device from cyberattacks, we need
to consider several cybersecurity levels. The first security level should be designed to
protect IoT sensor units, the second one has to shield the server where sensor data are
processed and control signals are issued to be delivered to actuators. Finally, the
third level is designed to protect the whole CPS application by monitoring trends
of cyberattack activities on the Internet.
When thinking about the first level of security, we should keep in mind that the
number of IoT units could be very large and heterogeneous in hardware and
software. Here, given the constrained hardware resources, it is mostly unrealistic
to implement a complex form of security such as installing antivirus software or
embedding a network intrusion detection mechanism in hardware. The protection
measures depend then on the computational resources of the end device. One
solution here would be to utilize an anomaly (fault) detection mechanism as those
used in fault diagnosis systems or an Oracle, as discussed in the Learning in
Nonstationary Environment section, earlier.