IEEE Consumer Electronics Magazine - November/December 2022 - 14

Securing the Future of Transporation
firewalls, or rule-based systems to detect cyberattacks.
These simple systems cannot detect
highly complex modern automotive attacks.
Another interesting trend in modern vehicles
Figure 1. Illustration of an in-vehicle network
connecting different ECUs using isolated networks
that are connected to a gateway, and external
connectivity of modern vehicles with various systems
in the environment.
Some of the most commonly used in-vehicle network
protocols include controller area network
(CAN), local interconnect network, FlexRay, and
Ethernet. Both ECUs and in-vehicle networks are
becoming more complex to satisfy emerging autonomyneeds.
Additionally,
a variety of automotive subsystems
rely heavily on the data from external systems,
as shown in Figure 1, which makes modern
vehicles highly vulnerable to various security
attacks. In the past decade (2010 onward), nearly
79.6% of all automotive attacks have been
remote attacks, which do not require the
attacker to be within the vicinity of the vehicle.1
A variety of attack vectors have been used,
including WiFi, telematics, Bluetooth, keyless
entry systems, and mobile applications. We discuss
many of these attacks in this article, as well
as techniques that have been proposed to protect
the vehicles from cyber-attacks. However,
due to the overall increase of the automotive
system complexity (heterogeneous ECUs, network
architectures/protocols, and applications),
detecting cyber-attacks is not easy, which poses
a major challenge for emerging connected and
autonomous vehicles (CAVs). There is a critical
need for a monitoring solution that can serve as
an intrusion detection system (IDS) to detect
cyber-attacks in vehicles. Traditionally, such
IDSs in computing systems have relied on using
14
is the widespread adoption of AI techniques for
advanced driver assistance subsystems (ADASs),
where environmental perception is required.2
Such AI techniques can also be deployed in powerful
automotive ECUs to monitor and detect
cyber-attacks. AI-based solutions are well-known
to be highly efficient in learning the complex patterns
that exist in high-dimensional time series
vehicular network data. They can observe for
anomalous patterns on in-vehicle networks that
connect all in-vehicle and external subsystems,
to detect cyber-attacks. With fully autonomous
vehicles supporting increased connectivity to
external subsystems on the horizon, having an
efficient IDS that can detect a variety of cyberattacks
using AI techniques is crucial and an
urgent requirement.
In this article, we discuss the timeline of major
automotive cyber-attacks and then present the
state-of-the-art efforts that utilize advanced AI
and deep-learning models to detect and address
cyber-attacks in vehicles. We also present a roadmap
for achieving cybersecurity goals in autonomous
vehicles and provide insights on the future
challenges in this area. This article is motivated
by the increasing frequency of attacks on automotive
platforms in recent years. The emphasis
on contemporary issues with cybersecurity in
autonomous vehicles provides valuable insights
for a general technical audience, as well as cyberphysical
architects, security researchers, automakers,
and Tier-N original equipment manufacturers
(OEMs) to understand the challenges in
this area and positively impact the design of
future security-aware automotive products.
HISTORY OF AUTOMOTIVE
CYBER-ATTACKS
Modern automotive systems are facing an
inflection point as they become highly vulnerable
to various security, privacy, and safety risks.
Several automotive attacks have been observed
in the past that ranged from targeting a single
stationary vehicle to a fleet of vehicles on the
road. Here, we present a timeline of the major
IEEE Consumer Electronics Magazine
IEEE Consumer Electronics Magazine - November/December 2022

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - November/December 2022
