Page 65 - Artificial Intelligence for the Internet of Everything

Intelligent Autonomous Things on the Battlefield 51

(deceptive information) produced by an adversary malware residing in
friendly things or otherwise inserted into the environment. Both humans
and things are susceptible to deception, and humans are likely to experience
cognitive challenges when surrounded by opaque things that might be
providing them with untrustworthy information (Kott & Alberts, 2017).
This situation reminds us that the adversarial nature of the battlefield
environment is a concern of exceptional importance, above all others.
The intelligent things will have to deal with an intelligent, capable adversary.
The adversary will bring about physical destruction, either by means such as
gunfire (also known as “kinetic” effects) or by using directed energy
weapons. The adversary will jam the communication channels between
things, and between things and humans. The adversary will deceive things
by presenting them with misleading information. Recent research in adver-
sarial learning comes to mind in this connection (Papernot et al., 2016).
Perhaps most dangerously, the adversary will attack intelligent things by
depositing malware on them.

3.3 AI WILL FIGHT THE CYBER ADVERSARY

A key assumption that must be made regarding the IoBT is that in a conflict
with a technically sophisticated adversary, IoBT will be a heavily contested
battlefield (Kott, Alberts, & Wang, 2015). Enemy software cyber agents,
or malware, will infiltrate our networks and attack our intelligent things.
To fight them, things will need artificial cyber hunters—intelligent, auton-
omous, mobile agents specialized in active cyber defense and residing
on IoBT.
Such agents will stealthily patrol the networks, detect the enemy mal-
ware while remaining concealed, and then destroy or degrade the enemy
malware. They will do so mostly autonomously, because human cyber
experts will be always scarce on the battlefield. They will be adaptive because
the enemy malware is constantly evolving. They will be stealthy because the
enemy malware will try to find and kill them. At this time, such capabilities
do not exist but are a topic of research (Theron et al., 2018). We will now
explore the desired characteristics of an intelligent autonomous agent
operating in the context of IoBT.
Under consideration is a thing—a simple senor or a complex military
vehicle—on which one or more computers resides. Each computer contrib-
utes considerably to the operation of the thing or systems installed on the