2
and isolation policies of each single remote control. Then
the attacker could abuse corresponding remote control in-
terfaces to hijack the TV. In fact, exploiting single remote
control does not result in severe security threats. However,
by combining functionalities of multiple remote controls,
one can design complex attacks.
To exploit an EVILSCREEN vulnerability, three consecu-
tive steps are needed. First, the attack utilizes less secure
wireless channels (i.e., IR and BLE) to enforce a Wi-Fi
provisioning [12], a common procedure for smart TVs to
receive credentials of a protected WLAN (i.e., SSID and
password). When inside the same WLAN, as most smart
TVs would not check the remote control pairing requests,
the attack leverages this weakness to actively bind a fake
remote control to the TV. Once the fake remote control is
bound to the TV, the attacker then abuses the remote control
interfaces to access TV resources and control the screen.
In comparison with attacks relying on meticulously
crafted signals (e.g., leveraging inaudible voice commands
to control the TV [13], [14], [15]), the EVILSCREEN attack
only uses common wireless technologies and is more gen-
eral. We conducted an empirical study against eight popular
smart TVs from retail markets of the China, Japan, Korea
and United States. Our study showed all of them were
vulnerable to the EVILSCREEN attack. Unlike attacks, such
as BIAS [16] against Bluetooth and KRACK [17] against Wi-
Fi, the EVILSCREEN attack does not aim to break any of the
three wireless protocols used by remote controls. Instead,
it exploits the fact that during communications between
the remote controls and the smart TV, because of usabil-
ity considerations, simplified security controls, or even no
security controls at all, are applied. We present a case
study for the Samsung smart TV, which adopts the ISO/IEC
30118-1:2018 standard [18] to protect its remote control
communication. We show that a usability factor related
to the Samsung SmartThings companion app significantly
reduced the crypto key randomness, and we constructed a
practical brute-force attack to breach its DTLS-over-BLE and
WebSocket-over-Wi-Fi communication between the TV and
its companion app simulated remote control.
The main contributions of this paper are as follows:
•New Understandings. We systematically analyzed
how the use of remote controls affects the security
of popular smart TVs. We show that design flaws
of remote controls break the security assumptions of
protection solutions currently deployed on wireless
technologies such as BLE and Wi-Fi.
•New Attacks. We implemented the EVILSCREEN at-
tack that affected 200 millions of popular smart TVs
worldwide 1. Unlike attacks aiming at exploiting
code vulnerabilities of TV OSes or apps, EVILSCREEN
attack only utilizes legitimate protocols and ser-
vices. Therefore, current protections are less effective
against our attack. We also outline countermeasures
for smart TV manufacturers and developers to miti-
gate the EVILSCREEN attack.
1. According to the shipments data reported by each smart TV
manufacturer [19].
2 BACKGROUND
In this section, we first give an analysis of smart TV char-
acteristics by comparing them with other three types of
devices. Then, we describe common protection schemes of
smart TVs, especially those to protect app simulated remote
control communications.
2.1 Characteristics of Smart TVs
Smart TVs provide a variety of new features and functions
for users, and thus their user experiences greatly differ from
other devices, such as smartphones and laptops. A TV is
considered “smart” when it has the following features: 1)
it relies on an OS to manage the hardware to process the
displayed contents; 2) it can access online media resources
through Internet connections; 3) it supports multiple ac-
cessories that communicate through various transmission
channels. Compared with other widely used electronic de-
vices including traditional TVs, smartphones as well as
common IoT devices, smart TVs have the following major
differences (see Table 1 for a summary):
Systems Because of the limited resources (e.g. small mem-
ory, limited power), traditional TVs and IoT devices are
usually not configured with a fully functional OS but just
with a tailored embedded OS, or even a bare metal firmware
with simple structures and functionalities. Smartphones,
with more powerful hardware, are equipped with OSes
(Android or iOS) and support different types of apps.
Like for smartphones, smart TV manufacturers often
customize OSes (TV OSes) to adapt to the smart TV hard-
ware and “smart” user interfaces. Most smart TV manufac-
turers build their TV OSes on top of an existing OS, such as
Android TV [9] developed by Google or tvOS [21] developed
by Apple. In addition, smart TV apps, like smartphone apps,
are provided to facilitate the use of smart TVs by users.
Smart TV apps are mostly provided as pre-installed apps by
smart TV manufacturers. Some of TV OSes, however, also
support the installation of third-party TV apps (often with
proprietary TV app stores).
User Interface (UI) Generally, a traditional TV consists of a
screen and a cable to display analog signals decoded media.
To support user interaction, traditional TVs usually display
menus on the screen for users to select. User interactive
inputs are limited to simple operations, such as switch-
on/off and channel/volume tuning, and such operations
are often conducted by the user with a remote control.
IoT devices, on the other hand, often lack a visible UI for
operations. Therefore, they usually rely on a remote web or
smartphone based user interface to handle user inputs. For
smartphones, the UIs are much more complex. With the help
of touchscreen, users can simply operate the smartphones
with different multi-touch gestures.
The UIs of smart TVs combine the features of UIs of
traditional TVs and smartphones. For usability consistency,
most smart TVs still support a menu based operation style,
but also support the use of TV apps (e.g., a media player
app) to enhance the functionalities as well.
Interaction Regarding the input styles, the interactions be-
tween users and the four types of devices differ significantly.
Traditionally, users operate the TV screen with a remote