# Facilitating Electromagnetic Side-Channel Analysis for IoT Investigation: Evaluating the EMvidence Framework

## Full Text

This Markdown is an automated text extraction for search and reading tools. The [hosted PDF](EvaluatingEMvidence.pdf) is the authoritative publication source.

Source PDF: [EvaluatingEMvidence.pdf](EvaluatingEMvidence.pdf)
Source SHA-256: `e7ccfd6eaee7c201744bd7ae60b4016e77ea9a16f0836e257400fdaf2e6903ce`
Generated with `pdftotext -nopgbrk`; formatting and reading order may differ from the PDF.

---

Forensic Science International: Digital Investigation 33 (2020) 301003

Contents lists available at ScienceDirect

Forensic Science International: Digital Investigation
journal homepage: www.elsevier.com/locate/fsidi

DFRWS 2020 USA d Proceedings of the Twentieth Annual DFRWS USA

Facilitating Electromagnetic Side-Channel Analysis for IoT
Investigation: Evaluating the EMvidence Framework
Asanka Sayakkara*, Nhien-An Le-Khac, Mark Scanlon
Forensics and Security Research Group, University College Dublin, Ireland

a r t i c l e i n f o

a b s t r a c t

Article history:

The Internet of Things (IoT) has opened up new opportunities for digital forensics by providing new
sources of evidence. However, acquiring data from IoT is not a straightforward task for multiple reasons
including the diversity of manufacturers, the lack of standard interfaces, the use of light-weight data
encryption, e.g. elliptic curve cryptography (ECC), etc. Electromagnetic side-channel analysis (EM-SCA)
has been proposed as a new approach to acquire forensically useful data from IoT devices. However,
performing successful EM-SCA attacks on IoT devices requires domain knowledge and specialised
equipment that are not available to most digital forensic investigators. This work presents the methodology behind and an evaluation of a framework, EMvidence, that enables forensic investigators to
acquire evidence from IoT devices through EM-SCA. This framework helps to automate and perform
electromagnetic side-channel evidence collection for forensic purposes. An evaluation of the framework
is performed by applying it to multiple realistic digital investigation scenarios. In the case of attacking
ECC cryptographic operations, the evaluation demonstrates that the volume of EM data that needs to be
stored and processed can be significantly reduced using the framework's machine learning based
approach.
© 2020 The Author(s). Published by Elsevier Ltd on behalf of DFRWS. All rights reserved. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords:
Digital forensics
Electromagnetic side-channels
Elliptic curve cryptography
Internet-of-things (IoT)
Machine learning

1. Introduction
Digital forensics involves data acquisition from digital devices in
order to help progress corporate, civil and legal investigations.
Traditionally, digital forensic investigators deal with personal
computers or mobile devices as the principal digital evidence
sources (Soltani and Seno, 2017). However, the emergence of the
Internet of Things (IoT) has revolutionised the potential for digital
forensics by opening up new sources of evidence (Lillis et al., 2016).
For example, wearable fitness tracker data can enable the precise
reconstruction of a person's movements. Similarly, smart home
data can reveal the exact time a person entered or left a premises.
While IoT devices can provide invaluable data for digital investigations, acquisition of data from IoT devices is not a straightforward task. An IoT device is a special purpose device designed to
perform a specific task. Several manufacturers produce these devices often with bespoke hardware and software designs. As a
result, IoT devices lack standard interfaces and forensic acquisition

* Corresponding author.
E-mail addresses: asanka.sayakkara@ucdconnect.ie (A. Sayakkara), an.lekhac@
ucd.ie (N.-A. Le-Khac), mark.scanlon@ucd.ie (M. Scanlon).

methods. This can often result in a device requiring a memory chipoff procedure in order to access its data (Watson and
Dehghantanha, 2016). However, with the increasing application
of lightweight cryptographic algorithms in IoT devices, such physical access into a device may not be a viable way to acquire forensic
data (Casey and Stellatos, 2008).
All running electronic devices emit electromagnetic (EM) noise
in various frequencies. This is due to the time-varying electrical
currents that are used. Similarly, running computers and mobile
devices are sources of strong EM noise. CPUs are considered to be
one of the strongest EM noise sources in computing due to fast
clock pulses used on them. The pattern of electrical pulses sent
through the CPU of a computer depends on the software instructions being executed and data being handled. Consequently,
CPU EM emissions has been shown to leak information about both
software activities and data (Getz and Moeckel, 1996).
EM side-channel analysis (EM-SCA) is a branch in information
security, which eavesdrops on these EM emissions (Kocher et al.,
1999). EM-SCA techniques have been used for various purposes
including software behaviour detection, software modification
detection, malicious software identification, and data extraction.
Among different data that can be extracted through EM-SCA

https://doi.org/10.1016/j.fsidi.2020.301003
2666-2817/© 2020 The Author(s). Published by Elsevier Ltd on behalf of DFRWS. All rights reserved. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

S2

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003

techniques, cryptographic keys are of significant forensic interest.
Various techniques have been developed for this purpose including
simple electromagnetic analysis (SEMA), differential electromagnetic analysis (DEMA) and correlation electromagnetic analysis
(CEMA).
The possibility of applying EM-SCA in digital forensic investigation scenarios involving IoT devices has been recently proposed
(Sayakkara et al., 2019a). When it is difficult or impossible to acquire forensic evidence from an IoT device, observing EM emissions
of the device can provide valuable information to an investigator.
Clues taken through EM-SCA may not be directly considered as
court-admissible digital evidence, they can still provide useful insights for an investigator to find court-admissible evidence elsewhere. However, EM-SCA requires specialised equipment and
technical expertise that many digital forensic investigators may
lack, which poses a considerable barrier to its adoption.
This work addresses the challenge of making EM-SCA a practical
reality to digital forensic investigators by outlining and evaluating a
software framework called EMvidence. A preliminary, proof-ofconcept of the framework is outlined in (Sayakkara et al., 2020a).
This work focuses on evaluating the EMvidence framework under
multiple application scenarios. The framework is designed to
facilitate extensiblity through an EM processing plug-in model.
Digital investigators can enable and use such extended capabilities
in practical cases, making a sustainable ecosystem.
Contributions of this work:

 Presentation of a methodology to extract forensically useful
insights from IoT devices through EM-SCA.

 Demonstration and evaluation of multiple IoT device investigation scenarios where an investigator can benefit by applying
EM-SCA methodology.

 Outline of an investigator-friendly open source software
framework that incorporates EM side-channel analysis capability for digital forensic purposes.

 Implementation and evaluation of a methodology to automatically separate EM traces of elliptic curve cryptography (ECC)
being performed on IoT devices, thus saving the storage space
required to store raw EM traces.

2. Background
2.1. Related work
Analysis of unintentional EM radiation has been identified as a
method to eavesdrop on electrical and electronic devices for decades (Sayakkara et al., 2019a). It has been demonstrated that
cathode ray tube (CRT) computer displays leak sufficient amount of
information to reconstruct the content being displayed through
their EM radiation (Van Eck, 1985). With the availability of off-theshelf hardware capable of capturing weak EM signals and computers with sufficient processing power to analyse data, EM sidechannel analysis on all kinds of computing devices became more
viable. As a result, a multitude of research has been conducted on
various EM-SCA techniques including software anomaly detection
and cryptographic key recovery (Sayakkara et al., 2019a).
Kocher et al. showed that power consumption of a computing
device can be used as a side-channel to extract cryptographic keys
(Kocher et al., 1999, 2011). When a CPU performs cryptographic
operations, each value assigned to its registers gets reflected in the
power consumption. By collecting a sufficient number of power
consumption traces during cryptographic operations with the same
key, it is possible to reveal the key using algorithms such as differential power analysis (DPA). Power consumption of the CPU is

directly associated with EM radiation of the device, which opens up
the EM side-channel. Therefore, variants of power analysis algorithms, such as DEMA, were introduced later in order to recover
cryptographic keys using EM traces (Quisquater and Samyde, 2001;
Gandolfi et al., 2001).
In addition to the information security aspects of EM-SCA
techniques, digital forensics is another field that can benefit from
EM-SCA techniques. Souvignet and Frinken suggested that power
analysis attacks can be used to extract data from smart-cards used
by malicious skimmer devices as a method to identify victims in a
forensic investigation (Souvignet and Frinken, 2013). However, it
requires physically tapping into the device being investigated
leading to potential tampering of evidences. In contrast, EM-SCA
techniques can be more suitable for digital evidence acquisition
as it does not require any physical alterations to the device being
investigated (Sayakkara et al., 2019b). However, cryptographic key
recovery is still a challenging task with EM-SCA due to the fact that
EM traces has to be acquired with precise alignment, which requires physical instrumentation of the device (Sayakkara et al.,
2018).
Carrying out EM-SCA attacks require precise acquisition and
analysis of EM traces. In order to ease the job of information system
security professionals to assess side-channel vulnerabilities of
embedded systems, it is necessary to have tools. ChipWhisperer
(ChipWhisperer embedded ha, 2019; O'Flynn and Chen, 2014) is a
widely used tool among security professionals and academic researchers to perform cryptographic key recovery attacks. It consists
of a collection of open-source trace acquisition hardware and data
analysis software components. Similarly, Riscure Inspector
(Riscure, 2019) is a fully fledged commercial product that comes
with software and hardware components to perform various power
and EM side-channel attacks to embedded devices including smartcards. While these tools are focused on information security objectives, we optimise the EMvidence framework for the specific
needs of digital forensic use cases.
2.2. Observation of electromagnetic side-channels
EM waves can be generated from electrical and electronic systems without the intention of the designers when conductors in a
circuit accidently behave as antennas (Maxwell, 1865; Jabbar and
Rahman, 1991). Electronic circuits that perform high-speed
switching operations are especially susceptible to generating unintentional EM noise due to their higher frequencies. Among them,
digital electronic components used on computers are well known
sources of EM noise since they employ high-speed clocks to carry
out their internal operations (Getz and Moeckel, 1996). CPU,
memory chips, data and address bus lines, various ports such as
USB and Ethernet are examples of EM radiation sources on a typical
computer system. Among these, EM radiation from the CPU is well
known to leak information about the internal activities of the CPU
including data being handled (Sayakkara et al., 2018).
When performing an EM-SCA attack, the information about
internal operations of the device being attacked are modulated into
the emission signal in various ways. This exact method of leaking
data through an EM side-channel is called the leakage model. It is
necessary to assume a specific leakage model when performing an
EM-SCA. From the inception of side-channel cryptographic key
recovery attacks, the major leakage model that has been explored is
Hamming weight leakage model. In multiple works by Kocher et al.
it has been shown that the Hamming weight of the data being
handled by a CPU gets modulated into the side-channel - hence the
name of the leakage model (Kocher et al., 1999, 2011). Another
closely associated model that often gets considered is the Hamming
distance leakage model where the number of bits that gets flipped is

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003

assumed to be modulated into the EM radiation (Brier et al., 2004).
Further improvements have even lead to the modelling of the exact
bit transitions, which can be either 0/1 or 1/0, called switching
distance leakage model (Peeters et al., 2007).

Aleak t 1⁄4 aHWðPt 4 Kt Þ þ ht

(1)

Fobserve 1⁄4 Fclock ±Fleak

(2)

When attacking a cryptographic algorithm with the intention of
retrieving the encryption key under Hamming weight model, it is
assumed that in a specific point in the execution of the algorithm,
the Hamming weight of the cryptographic key bits are exposed
(Robyns et al., 2019). For example, consider a simple cipher where a
plaintext, P, is associated with a key, K, through XOR operations to
generate the ciphertext. The amplitude of the information leaking
EM signal Aleak t can be modelled with the Hamming weight
leakage model as shown Equation (1). HW is the Hamming weight
function while Pt and Kt are the plaintext and key bytes XOR-ed at
the time instance t. a is a positive integer used as a scaling factor
while ht is the noise at time t. When a signal gets modulated with a
carrier wave such as CPU clock or on-board radio transmitter signal
through the amplitude, it causes side-bands to occur between the
carrier wave (Camurati et al., 2018). For example, consider the clock
frequency of a CPU to be Fclock and the frequency of the leakage
signal to be Fleak . This causes the observation of a frequency
bandwidth Fobserve that spans from ðFclock Fleak Þ to ðFclock þFleak Þ as
illustrated in Equation (2).
Observation of unintentional EM emissions from computing
devices can be made using traditional signal analysis hardware,
such as oscilloscopes and spectrum analysers. As an alternative,
software defined radios (SDR) where fast analogue-to-digital converters (ADC) are used to digitise EM signals and feed into software
for processing and visualisation. Compared to the traditional options, SDRs provide more flexibility and ease of use to non-signal
analysis professionals. Fig. 1 illustrates the EM signals observed at
clock frequencies of two representative IoT devices using an SDR
device setup, i.e., a Raspberry Pi (1:4GHz CPU) and an Arduino
Leonardo (running at 16MHz). In both cases, an H-loop antenna is
placed slightly above the processor chip of the devices. The antenna
is connected to a HackRF SDR device (Ossmann, 2016) that is finally
connected to a host computer running GNURadio software library
(Blossom, 2004) to process data. The frequency region of the
Arduino was too noisy to observe directly. Therefore, it was
empirically identified that the 18th harmonic of the clock

Fig. 1. Leakage signals of two representative IoT devices - (a) Raspberry Pi 3 Bþ at
1:4GHz and (b) Arduino Leonardo at 288MHz (18th harmonic).

S3

frequency, i.e., 288MHz, is most appropriate to observe EM emissions of the Arduino device.
3. EMvidence forensic framework
This section presents the design and functionality of the EMvidence software framework for EM-SCA for digital forensic purposes.
The key principle behind the design of EMvidence framework is to
make it easy to use for digital forensic investigators who are
generally non-experts of EM-SCA. Furthermore, due to the rapid
changes that occur in IoT device ecosystem, it is intended to make
the framework easily extendable by third parties - especially with
new ML models to support new IoT devices and new information
gathering.
3.1. High-level architecture
In order to gather forensically useful information using EM-SCA
techniques, a software tool should facilitate three main phases. The
first is the training/research phase where IoT devices and their
forensically useful software activities are profiled and added into
the framework. Once the framework with such analysis capability
is ready, it can be used in the second phased called the field
investigation phase. There, an investigator can collect EM data from
a suspect IoT device, referred to as ‘‘device under test’’ (DUT), in a
real-world scenario and gather insights about the device on-thespot. The third phase is further analysis where the device is taken
into a forensic laboratory and used to perform advanced EM-SCA
methods such as cryptographic key recovery attacks. Fig. 2 illustrates these three phases of EM-SCA based digital forensics.
EMvidence framework consists of a main core with multiple
default modules and facility to add third-party plug-ins depending
on future requirements. Main component of EMvidence is its core
GUI that provides the default interface to a user. It also manages the
modules and plug-ins by establishing communication between
them in a coordinated fashion. Together with core GUI, the
framework comes with three default software modules that are
essential to the normal operations of the framework namely, data
acquisition module, data visualisation module, and report generation module. Furthermore, depending on the requirements, thirdparty users can develop and add plug-ins to the core GUI of the
EMvidence framework. Such plug-ins may provide various data
analysis capabilities such as software behaviour detection, cryptographic key recovery, etc. The source code of the EMvidence

Fig. 2. The three phases of EMvidence framework usage.

S4

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003

Fig. 3. Controlled/Instrumented EM signal Acquisition.

framework and it's associated plug-ins are available at a Github
repository.1

represents amplitude. In order to see the individual frequency
components of a signal, it can be converted to the frequency
domain using Fast Fourier Transform (FFT). Furthermore, FFT can be
applied in short time intervals, i.e., Short-Time Fourier Transform
(STFT), to generate a spectrogram e a graph where the x-axis
represents observation time and the y-axis represents the frequency. The colour codes are used within a spectrogram to represent the amplitude of the signal at a particular time instance in a
particular frequency.
There are two methods for EM data visualisation. The first is
reading data from the SDR device and real-time visualisation
thereof. This approach is useful for visually identifying the frequencies where information leakages occur from the DUT. A spectrogram is generally used to observe such suspicious EM signals for
information leaking patterns. The second method is visualising EM
data taken from saved files. Such off-line EM data visualisations are
useful for EM traces acquired outside of EMvidence. Users can
interpret such an externally-captured EM trace file by observing it
visually with EMvidence. An FFT can be used in this scenario (see
Fig. 4).

3.2. Data acquisition module
This module facilitates the acquisition of EM data for analysis.
For this purpose, a SDR device needs to be connected to the host
computer where the EMvidence framework is running. Since the
framework makes use of GNURadio software library (Blossom,
2004) to handle SDR interfaces, any SDR device supported by
GNURadio can be used. This module supports two types of data
acquisition methods. Firstly, observation of EM emission signal can
be made for a predefined period of time without any interaction or
communication with the DUT from few centimetres away from it.
This is the approach used in a digital forensic investigation scenario.
Secondly, EM signals can be acquired while actively interacting
with the DUT in scenarios where it is safe to communicate with the
device through an interface such as universal serial bus (USB),
universal asynchronous receiver/transmitter (UART), or Ethernet.
While interactive EM data acquisition can be used to profile a
new type of IoT, it requires precise coordination between the SDR,
DUT, and the host computer. Consider a scenario where it is
required to build a ML model to detect a specific piece of software
code running on an IoT processor/microcontroller. As the training
samples for ML, individual EM traces are needed e each representing the exact time period where particular software is
executed. EMvidence provides functionality to send commands to
the target device to start and stop performing a particular task
through USB or UART interfaces. Users can program target devices
to run associated programs upon receiving these commands. With
this setup, EMvidence can command the target device to start the
required software operation and start saving EM data from the SDR
device. Once EMvidence receives feedback that software operation
is complete, it immediately ceases data acquisition. By repeating
this process, a large number of EM trace files can be acquired
representing a specific software behaviour on the target device.
Fig. 3 illustrates the acquisition of EM traces in a coordinated
fashion from a target device. Once such data are used to build ML
models, they can be incorporated into EMvidence framework to
inspect similar IoT devices in investigative scenarios.

3.4. Third-party plug-ins
The IoT ecosystem is highly dynamic in nature. New IoT devices
enter into the market continuously while existing IoT devices can
change frequently due to software updates. Furthermore, various
novel EM-SCA analysis techniques are released time to time that
can be used to inspect IoT devices. Keeping up with these changes is
not an easy task for any digital forensic investigation tool. This is
where the necessity arises to support development of third-party
plug-ins to the software tool depending on user requirements.
When a large community of users continuously develop and provide new plug-ins to the platform, it can keep up with the dynamic
IoT ecosystem.
EMvidence plug-ins enable extra functionality to analyse EM
data. EMvidence provides an application programming interface
(API) for plug-ins to take services from the core framework, such as
providing access to real-time EM signal samples or saved EM trace
files. Furthermore, a plug-in can provide information back to the
core framework. Consider a situation where a user develops a new
ML model to recognise a particular software activity running on a
specific IoT device. This trained model can be easily integrated into
EMvidence by wrapping it with API calls to the framework. Once
developed, such plug-ins can be either distributed and used among
the users of EMvidence framework or submitted to the source code

3.3. Data visualisation module
In the simplest form, an EM signal can be visualised as a time
domain signal; where the x-axis represents time and the y-axis

1

https://github.com/asanka-code/EMvidence.

Fig. 4. Power Spectral Density (PSD) of the IoT device's EM signals.

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003

repository of the framework in order to be distributed officially.
When an investigator enables this particular plug-in during an
investigation, EMvidence delivers EM signal data to the plug-in, lets
the investigator interact with it, and receive ML classification results back.
4. Evaluation
The goal of this section is to evaluate the potential of using
EMvidence framework in multiple practical aspects of information
gathering from IoT devices. IoT devices that are in wide use in dayto-day life comes in all shapes and sizes. While each IoT device can
consist of a unique combination of hardware and software, they can
broadly be categorised into two classes based on their computing
resources as low-end and high-end IoT devices. The design choice of
choosing computing hardware resources to be included in an IoT
device depends on multiple reasons. Among them, source of power
is an important factor. A device that can be continuously mains
powered can have more powerful hardware, and therefore, can be
considered as a high-end IoT device. Meanwhile, a device that has
to rely on a battery for a prolonged period of time needs to use an
energy efficient processor such as micro-controllers or system-onchips (SoC). Hence, such devices can be considered as low-end IoT
devices and include health implants, fitness wearable devices, and
smart light bulbs. Due to limited on-board storage, these devices
generally do not store much data. They tend to either transmit data
into an associated smart-phone app or hub, or directly to a cloud
service. Therefore, even if a low-end IoT device contains some nonvolatile data storage, e.g., an SD card, it is less likely that traditional
digital evidence extraction methods would prove fruitful.
Meffert et al. highlighted the need for identifying the running
forensic state of IoT devices in an investigation (Meffert et al., 2017).
A forensic state is the state of hardware and software of an IoT
device at the time it was seized by law enforcement. For example,
an IoT smart lock can have two states, i.e., locked and unlocked, and
its state could be a vital information in an investigation. Furthermore, IoT devices that are subject to investigation may have been
tampered with intentionally or as a result of malware. Ronen et al.
demonstrated that IoT smart bulbs can be infected over-the-air
(OTA) and be controlled remotely (Ronen et al., 2017). Such maliciously modified devices are shown to be effective in causing
harmful results to humans, such as creating epileptic seizures to
vulnerable individuals by adjusting the frequency of LED smart
bulbs (Ronen and Shamir, 2016). Therefore, verifying whether the
device is running its default firmware can useful. In case the device
has been reprogrammed and protected with encryption, identifying the encryption algorithm is also forensically useful.
4.1. Processing electromagnetic trace data
EM-SCA attacks require a sufficient number of target device EM
traces. Furthermore, in order to train ML models using EM traces, it
is necessary to have EM trace samples annotated with the specific
software activities of the target device they represent. Three
hardware components are necessary for the acquisition of EM
traces. These are namely; a DUT, a signal capturing device, and a
host computer. The signal capturing device is connected to the host
computer via USB interface while the DUT may or may not be
connected in a similar manner. The host computer runs EMvidence
framework and stores the captured EM traces for analysis.
An EM trace is a vector that represents the amplitude variation
of a signal over time. Due to fast sampling rates used by signal
acquisition hardware, an EM trace with a duration of milliseconds
can contain millions of data points. When these data are used
directly as input to train and test machine learning and deep

S5

learning models, the highly dimensional data can negatively affect
time and amount of computing resources they demand. As a result,
EM traces acquired through the aforementioned hardware setup
are not suitable to be directly used to train DL models. Therefore,
EM traces are pre-processed in order to transform them from a
continuous time domain signal into a format that has a manageable
feature vector for DL. LSTM architecture was used for the deep
neural network, which is suitable for the identification of patterns
that occur in time series data, such as EM traces (Sayakkara et al.,
2020b).
When attempting to classify software activity EM traces,
labelled EM traces are needed. For this purpose, EM trace samples
are acquired by running each software activity on the target device
and collecting EM traces of a predefined length. Originally, each EM
trace is in time domain. Time-domain signals are prone to fluctuations caused by external noise. Therefore, each trace is transferred
to frequency domain using FFT with an overlapping sliding window. For each EM trace, this results in a collection of FFT vectors
representing consecutive time steps. The dimensions of the FFT
vectors are still considerably higher to be directly used as the
feature vector for LSTM classifier, e.g., 200; 000 dimensions.
Therefore, the dimensions of these FFT vectors are further reduced
by dividing the elements of each FFT vector into 1; 000 equally
spaced buckets. From each bucket, the maximum element is
selected as the representative of the bucket without losing the
generalisation. This results in a 1; 000 element long feature vector
for each time step of EM traces.
4.2. Forensic state detection
In order to illustrate the usefulness of detecting forensic state of
low-end IoT devices through EM-SCA, the following hypothetical
scenario was considered. An IoT device has been deployed in a
building as a part of an intruder detection system. The device
consists of a sensor that detects movements within a specified
space of the premises. The device consists of two actuators, an
alarm and a door lock, that it can control independently. Furthermore, the device is connected to a GSM module in order to send and
receive SMS. The device firmware is programmed to continuously
read the motion sensor to detect intrusions into the premises. Upon
detection, it can perform one of three tasks e locking the door,
firing an alarm, or sending a text message to the owner. At any time,
the device can be disabled by pressing a physical button that puts
the device into an idle state. The device does not have any other
associated network servers that can log device states. Furthermore,
the device does not switch internal states due to any other reason
than the specified ones. The five states of the device that we are
interested in are namely; (1) idle, (2) read digital sensor (reading
motion sensor), (3) control digital actuator (firing the alarm), (4)
control analogue actuator (turning door lock), and (5) serial
communication (sending text message).
Suppose that this building is subject to a legal investigation for a
crime - assumed to be conducted by an intruder. Upon the arrival of
law enforcement, one of the investigative questions is whether the
intruder detection system functioned as expected or did an insider
disable it before the crime occurred. The answer to this question
can be found if the current internal state of the device is known.
Turning the device off and moving it to a digital forensics laboratory
destroys the current internal state of the device. Performing a live
EM-SCA on the intrusion detection device and identifying its current software state may be the only viable approach to resolve this
problem.
The IoT device was emulated by using an Arduino Leonardo device. It was programmed to run a software code that puts the device
on each of the 5 states chosen by the user. While the device was

S6

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003
Table 1
Private and public key sizes of ECC curves.
Curve

Private Key (bytes)

Public Key (bytes)

secp160r1
secp192r1
secp224r1
secp256r1
secp256k1

21
24
28
32
32

40
48
56
64
64

Fig. 5. Confusion matrix of the IoT device state classifier.

running on each state, a 30-s long EM trace was captured per state
with a sampling rate of 20MHz using HackRF SDR. The SDR was
tuned into the 18th harmonic of the Arduino's clock frequency, i.e.,
288Mz. The H-loop antenna of the SDR was placed over the processor of the device during data acquisition. Fig. 4 illustrates the
power spectral density (PSD) of the EM signals from IoT device in its
different states. A neural network classifier based on multi-layer
perceptron (MLP) architecture was selected to distinguish each
state of the IoT device. A non-overlapping sliding window with a
width of 250 ms was used to extract EM trace segments; subsequently converted to the frequency domain. These windows were
grouped and averaged to produce a vector of 1; 000 features that
were considered as training and testing samples for the ML
classifier.
Fig. 5 illustrates the confusion matrix of the classification results.
The classifier was able to achieve an average F1-score of 99% in
distinguishing the 5 IoT device states. This indicates that a pretrained model to identify internal software states of the IoT device. For example, if it was identified that the device is in the idle
state at the time investigators arrived at the scene, it's clear that
someone deliberately turned the device into idle state in order to
stop it from triggering the intruder alarm. In that case, fingerprints
on the button of the IoT device could potentially help to identify the
insider. Once the machine learning classifier was built, it is integrated into the EMvidence as a third-party machine learning model
for identifying internal state of the particular type of IoT devices.
4.3. Elliptic curve detection
Some cryptographic algorithms, e.g., RSA, demand reasonably
high computational power making them unsuitable for lowpowered computing devices. As a result, elliptic curve cryptography (ECC) has increasingly been deployed on these devices. ECC is
a public key cryptography that requires a smaller key length
compared to RSA. Numerous different elliptic curves can be used in
ECC. Table 1 illustrates five major elliptic curves designed to run on
low power devices (available in the micro-ecc library). These elliptic
curves use different private and public key lengths with bespoke
configuration settings.
Due to the differences in settings of each elliptic curve, the
running time of ECC operations for each curve can vary. Fig. 6 illustrates the average time different elliptic curves take to digitally
sign a message and to verify the signature of a message (ECDSA).
The measurements were calculated by running each ECDSA algorithm on an Arduino device with equally sized messages. For each

Fig. 6. The time it takes to digitally sign and verify a message using different ECC
curves on an Arduino device.

curve, both signing and signature verifying operations take almost
the same amount of average time. However, the average time taken
by individual curves are distinguishably different from each other.
In order to investigate the possibility of detecting the presence
of ECC cryptography through EM emissions, an experiment was
conducted with an LSTM binary classifier. The objective is to train a
model to distinguish between ECC cryptography operations and
other software activities. For this purpose, an Arduino device was
programmed to perform ECDSA signing operations that was
controlled by sending commands through USB from the host
computer. Each time an ECDSA signing operation is performed, a
100 ms trace was captured through a H-loop antenna placed over
the Arduino's microcontroller chip connected to a HackRF software
defined radio.
For each of the five elliptic curves, 50 EM traces were acquired
(totalling 250 traces) for the ECC cryptography class. For non-ECC
cryptography operations class, 20 Arduino programs used that
have varying complexities. From each Arduino program, 12 traces
were acquired for non-ECC cryptography class (totalling 240
traces). A sliding window of 10 ms was used with a 2 ms step size
(80% overlap) to collect segments from each trace and subsequently
a feature vector for each window segment was calculated using FFT
broken down into 1; 000 equally spaced buckets. Each bucket's
maximum amplitude frequency component was selected as the
representative signal for the bucket. This results in training sequences each having time steps where each time step consists of
1; 000 features. All the training samples were normalised to values
between 0 and 1. Fig. 7 illustrates examples of signals acquired from
a DUT while running two curves of ECC and running two non-ECC
programs.
The LSTM classifier was implemented using Keras library with
python. A single LSTM layer consists of 100 nodes and a fully connected layer with 1 node for binary classification. This last node
uses a sigmoid activation function, while the model uses binary
crossentropy as the loss function. The sequence data set was broken

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003

Fig. 7. Example ECC and non-ECC Signals Acquired.

into 75% and 25% sets for training and testing purposes respectively.
When using 5 epochs and 64 batch size, the LSTM classifier achieved an impressive 100% accuracy. In order to assess the effect
from the sliding window length and EM trace length, further LSTM
models were trained and tested varying those parameters. Fig. 8
illustrates the variation of classification accuracy against both
sliding window length and EM trace length. As is evident, the
longer the EM traces, the higher the classification accuracy achieved. This is due to the fact that longer EM traces results in longer
sequences with more information for the LSTM model to learn. In
contrast, longer sliding window lengths negatively affected the
classification accuracy, reducing it to 95% in the worst case. Again,
the reason is behind the length of the sequences. Longer windows
result in shorter sequences for a fixed length of EM traces.

4.4. Automated trace segmentation
When an IoT device is employing data encryption, EM-SCA
techniques can assist an investigator to retrieve data decryption
key. However, the success of cryptographic key recovery techniques
depends on the number of EM traces acquired by observing a target
device during cryptographic operations. These EM traces need to be
acquired in a manner such that the cryptographic operation being
observed must occur at the same point in time across all the traces.
It has been shown that the accuracy of key recovery is considerably
affected by misalignment of the EM traces (Kocher et al., 2011). The

most common approach to extract sufficiently aligned EM traces is
by instrumenting the target device through software or hardware.
The purpose of this instrumentation is to identify the beginning
and ending time where the cryptographic operation occurs so that
the EM observation hardware can be synchronised to extract EM
traces within that precise time period.
Instrumentation of target devices can be used to demonstrate
key recovery attacks, but it is not viable in most real-world application scenarios. Such scenarios occur when the attacker has no
physical access to the device being attacked or when implementing
such instrumentation is not possible. Under such circumstances,
the EM traces have to be gathered without instrumentation and
later be processed in order to realign them. Such post processing is
a computationally costly task where the success of alignment is not
sufficiently guaranteed.
When an Arduino device is performing a cryptographic operation, the experimental setup captures the data over a bandwidth of
20MHz centred at 288MHz. Fig. 9 illustrates the spectrogram of a
signal that was observed while the device was performing two
consecutive ECC digital signature operations. As can be seen, two
distinctive regions across the time dimension have changed patterns corresponding to the two ECC operations. However, not all
captured bandwidth leaks information about software activities. It
is evident that some channels represent significant changes in
signal amplitude compared to others in the two interested regions.
Fig. 10 illustrates two selected channels from the captured bandwidth where the first channel clearly represents two unique areas
corresponding to two ECC operations and the second channel
represents only a common region of change in pattern. This means
that finding the right channel can help to identify the precise time
instance where a particular software task has started and ended.
The automated separation of EM traces relevant to ECC operations was performed as follows. The target IoT device, i.e., Arduino,
was programmed to run an ECC-related triggered by a command
sent through USB. The exact task it excutes is using a private key
generated using curve secp160r1 to digitally sign a 32 byte long
arbitrary message. The controlling host computer sends a command every 5 s in order to make the desired operation occur
periodically at fixed intervals. Meanwhile, EM traces were acquired
by observing for a predefined time period of 30 s. For such captured
EM traces, a STFT was applied with a window size of 1 ms. Due to
the sampling rate of 20 MHz, the STFT operation results in a spectrum dataset with 20; 000 channels. Finally, one manually selected
channel from the EM data is applied to a change point detection
algorithm, The Pruned Exact Linear Time (PELT) (Killick et al., 2012),
in order to identify time instances where the significant pattern
changes occur. Once the change points of an EM trace were

Fig. 9. Signal spectrogram of ECC digital signing two times.
Fig. 8. Ecc digitally signing and verification time on an Arduino.

S7

S8

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003

Fig. 10. Two Arduino EM channels when digitally signing with ECC.

identified, the trace was broken into it corresponding segments.
Finally, a ML model trained to identify ECC curves was used to
distinguish those EM trace segments with ECC operations. Keeping
the EM segments of ECC operations for future use, such as cryptographic key recovery attacks, the rest are discarded.

4.5. Processing overhead
High sampling rates are necessary to extract as much CPU data
leakage information as possible. It is necessary to observe EM signals for a longer time duration in order to gather sufficient traces to
attack cryptographic operations. However, on the down side, these
two factors considerably increase the size of the resultant EM trace
files saved on the host computer. A significant percentage of such
saved data may not contain intended software activities, wasting
storage space and processing time. As a result, it is useful to process
EM signals in real-time and save traces only when activity of interest is detected.
The overhead of processing EM signals in real-time is evaluated
as follows. While EM signal capture device is set to 20 MHz sampling rate, a sliding window with a fixed width of 10 ms was used to
slide through the real-time I-Q data feed. Each data window was
then preprocessed in real-time to generate the features and fed to a
neural network-based binary classifier to detect the presence of
ECC cryptography operations. The step size of the sliding window,
i.e., the amount of overlap between consecutive windows, was
varied between 0.5 ms and 10 ms for different independent trials.
In all experiments, the total signal capturing duration was fixed to
10 s.
In Fig. 11, the graph on the top illustrates the number of windows produced against the sliding window step size. When
reducing step size, the number of windows available to process for
a given time period increases exponentially. The graph at the bottom of Fig. 11 illustrates the statistics of the number of data samples
waiting in the real-time buffer until the sliding window had processed them. As it can be seen, the median length of the I-Q signal
buffer does not indicate any noticeable increase even for the
smallest sliding window step size considered. This means, even
though the number of sliding windows to process increases with
the reduction of sliding window step size, it seems not to incur any
considerable overhead to the real-time processing buffer. The
production and consumption of the EM samples were in an
equilibrium.

Fig. 11. Effect of the sliding window step size to the data collection buffer over 10 ms.

5. Discussion
In traditional digital forensics, the data acquired from devices
under investigation are handled in a forensically sound way in order to make sure the court admissibility of evidence. For example,
when a disk image is acquired from a computer, cryptographic hash
values are calculated and stored along with the disk image. The
hash values can later be used to verify the integrity of an image.
Similarly, EM traces acquired from IoT devices has to be stored with
a hash verification facility. The EMvidence framework supports
hash calculation for EM traces acquired in real-time and stores
them along with the traces. However, the patterns of the EM signals
from an IoT device depends on its current internal states and
external noise sources in the vicinity. Therefore, hash values are
useful only to maintain the integrity of the originally acquired EM
traces during subsequent analysis. Further steps to ensure forensic
soundness of EM data acquisition and analysis require more studies
that we hope to conduct in the future.
The experimental evaluations with the EMvidence framework
indicate that EM-SCA based inspection of IoT devices can be useful
for forensic investigations in a variety of scenarios. Detection of the
forensic state, specific ECC curve, and automatically segmenting
ECC-related EM traces are device-specific tasks. Performing each of
those tasks with IoT devices by using tailor-made ML models requires profiling of each individual IoT device of interest. This cannot
be done without the support of a large community due to the dynamics of IoT ecosystem. This need reinforces the necessity of opensource and extendable platforms like EMvidence.
The focus of EM-SCA techniques is to consider direct unintentional emissions from the processor as a side-channel. However,
there can be useful alternative side-channels that are also highly
relevant. For example, IoT devices are often equipped with Systemon-a-Chip (SoC) processors that contains a CPU and a radio transceiver on the same chip. Due to the close proximity between the
CPU (a digital component) and the radio transceiver (an analogue
component), it has been shown that the CPU's digital operations
can affect the analogue circuitry. This causes information leakage
from the CPU to the on-chip radio transceiver (Camurati et al.,
2018). Therefore it is possible to observe the transmissions of the
radio from longer distances and extract CPU operation-related

A. Sayakkara et al. / Forensic Science International: Digital Investigation 33 (2020) 301003

information. This is a less explored opportunity that can have
digital forensic use cases, such as when a device needs to be
inspected from a distance.
If the data stored on the IoT device are not encrypted and the
device has a standard interface, the data can be extracted using
existing forensic evidence acquisition methods (MacDermott et al.,
2019). However, many IoT devices lack these standard interfaces,
often forcing investigators to take more risky approaches, such as
chip-offs. Mistakes during such operations could inadvertently
destroy useful evidence on a device. Therefore, in cases where a
chip-off is being considered, it may be a good approach to first try
an EM-SCA inspection on the device to gather as much information
as possible before attempting a chip-off.
There are various hardware and software mitigation techniques
available to counter EM information leakage (Zankl et al., 2018). It is
possible that an IoT device may have applied EM side-channel
mitigation techniques into its firmware in order to mislead EMSCA attacks. It is possible that such mitigation techniques could
cause ML classifiers to return inaccurate classification results.
Further studies are required to assess the impact of such mitigation
approaches to the EMvidence framework. However, due to the
computational and energy cost of such mitigation techniques, they
are rarely applied in low-powered devices making them less
immediately threatening.
6. Conclusion and future work
With the ever-increasing applications of IoT systems in domestic
and industrial environments, digital forensic investigations
increasingly require the extraction of digital evidence from them.
Most forensically useful information in IoT devices are currently
extracted by intrusive inspections of hardware that makes them
less forensically sound. This work presented the design of EMvidence, a framework for digital forensic investigators and researchers to leverage unintentional EM radiation from IoT devices
as an information source. EMvidence is designed in a manner that it
can be easily extended with new functionalities to keep up with the
dynamism of IoT devices. Experimental demonstrations proved
that ML classifiers can be used to gain useful insights in IoT
investigative scenarios.
Several future work directions exist. The experimental demonstrations provided in this paper used an Arduino and a Raspberry Pi
as representative IoT devices. It is necessary to evaluate EMvidence
with the most commonly encountered IoT devices in real-world
digital forensic scenarios. Currently, the framework is tested only
with HackRF SDR as the EM signal acquisition device. It is necessary
to test the inseparability between different SDR devices. For
example, a ML model trained from EM traces from one specific SDR
device must be accurate in classifying EM signals captured with
other SDR devices.
References
Blossom, E., 2004. GNU radio: tools for exploring the radio frequency spectrum.
Linux J. 2004 (122), 4.
Brier, E., Clavier, C., Olivier, F., 2004. Correlation power analysis with a leakage
model. In: International Workshop on Cryptographic Hardware and Embedded
Systems. Springer, pp. 16e29.
Camurati, G., Poeplau, S., Muench, M., Hayes, T., Francillon, A., 2018. Screaming
channels: when electromagnetic side channels meet radio transceivers. In:
Proceedings of the 25th ACM Conference on Computer and Communications
Security (CCS). CCS ’18. ACM.
Casey, E., Stellatos, G.J., 2008. The impact of full disk encryption on digital forensics.
ACM SIGOPS - Oper. Syst. Rev. 42 (3), 93e98.
ChipWhisperer Embedded Hardware Security Toolchain, 2019. Accessed. https://
newae.com/tools/chipwhisperer/. (Accessed 1 October 2019).
Gandolfi, K., Mourtel, C., Olivier, F., 2001. Electromagnetic analysis: Concrete results.

S9

In: International Workshop on Cryptographic Hardware and Embedded Systems (CHES). Springer, pp. 251e261.
Getz, R., Moeckel, B., 1996. Understanding and Eliminating Emi in Microcontroller
Applications. National Semiconductor.
Jabbar, M., Rahman, M.A., 1991. Radio frequency interference of electric motors and
associated controls. IEEE Trans. Ind. Appl. 27 (1), 27e31.
Killick, R., Fearnhead, P., Eckley, I.A., 2012. Optimal detection of changepoints with a
linear computational cost. J. Am. Stat. Assoc. 107 (500), 1590e1598.
Kocher, P., Jaffe, J., Jun, B., 1999. Differential power analysis. In: Advances in Cryptology (CRYPTO ‘99). Springer, 789e789.
Kocher, P., Jaffe, J., Jun, B., Rohatgi, P., 2011. Introduction to differential power
analysis. Journal of Cryptographic Engineering 1 (1), 5e27.
Lillis, D., Becker, B., O'Sullivan, T., Scanlon, M., 2016. Current challenges and future
research areas for digital forensic investigation. In: The 11th ADFSL Conference
on Digital Forensics, Security and Law (CDFSL 2016), vols. 9e20. ADFSL, Daytona
Beach, FL, USA.
 Lea, S., Iqbal, F., Idowu, I., Shah, B., 2019. Forensic analysis of
MacDermott, A.,
wearable devices: Fitbit, garmin and hetp watches. In: 2019 10th IFIP International Conference on New Technologies, Mobility and Security (NTMS). IEEE,
pp. 1e6.
Maxwell, J.C., 1865. A dynamical theory of the electromagnetic field. Phil. Trans. Roy.
Soc. Lond. 155, 459e512.
Meffert, C., Clark, D., Baggili, I., Breitinger, F., 2017. Forensic State Acquisition from
Internet of Things (FSAIoT): a general framework and practical approach for IoT
forensics through IoT device state acquisition. In: Proceedings of the 12th International Conference on Availability, Reliability and Security. ACM, p. 56.
Ossmann, M., 2016. Software defined radio with hackrf. Great Scott Gadgets https://
greatscottgadgets com/sdr.
O'Flynn, C., Chen, Z.D., 2014. ChipWhisperer: an open-source platform for hardware
embedded security research. In: International Workshop on Constructive SideChannel Analysis and Secure Design. Springer, pp. 243e260.
Peeters, E., Standaert, F.X., Quisquater, J.J., 2007. Power and electromagnetic analysis: Improved model, consequences and comparisons. Integrat. VLSI J. 40 (1),
52e60.
Quisquater, J.J., Samyde, D., 2001. Electromagnetic Analysis (EMA): Measures and
Counter-measures for Smart Cards. Smart Card Programming and Security,
pp. 200e210.
Riscure, 2019. Inspector SCA: Side-Channel Analysis Tool for Embedded Systems.
Accessed. https://www.riscure.com/security-tools/inspector-sca/. (Accessed 1
October 2019).
Robyns, P., Quax, P., Lamotte, W., 2019. Improving cema using correlation optimization. IACR Transactions on Cryptographic Hardware and Embedded Systems
1e24.
Ronen, E., Shamir, A., 2016. Extended functionality attacks on IoT devices: the case
of smart lights. In: 2016 IEEE European Symposium on Security and Privacy
(EuroS&P). IEEE, pp. 3e12.
Ronen, E., Shamir, A., Weingarten, A.O., O'Flynn, C., 2017. IoT goes nuclear: creating a
ZigBee chain reaction. In: 2017 IEEE Symposium on Security and Privacy (SP).
IEEE, pp. 195e212.
Sayakkara, A., Le-Khac, N.A., Scanlon, M., 2018. Electromagnetic side-channel attacks: potential for progressing hindered digital forensic analysis. In: Companion Proceedings for the ISSTA/ECOOP 2018 Workshops. ISSTA ’18. Association
for Computing Machinery, New York, NY, USA, ISBN 9781450359399,
pp.
138e143.
https://doi.org/10.1145/3236454.3236512
doi:10.1145/
3236454.3236512.
Sayakkara, A., Le-Khac, N.A., Scanlon, M., 2019a. A survey of electromagnetic sidechannel attacks and discussion on their case-progressing potential for digital
forensics. Digit. Invest. 29, 43e54. https://doi.org/10.1016/j.diin.2019.03.002.
URL: http://www.sciencedirect.com/science/article/pii/S1742287618303840.
Sayakkara, A., Le-Khac, N.A., Scanlon, M., 2019b. Leveraging electromagnetic sidechannel analysis for the investigation of IoT devices. Digit. Invest. 29,
S94eS103.
https://doi.org/10.1016/j.diin.2019.04.012.
URL:
http://www.
sciencedirect.com/science/article/pii/S1742287619301616.
Sayakkara, A., Le-Khac, N.A., Scanlon, M., 2020a. EMvidence: a framework for digital
evidence acquisition from IoT devices through electromagnetic side-channel
analysis. Forensic Sci. Int.: Digit. Invest. 32 https://doi.org/10.1016/
j.fsidi.2020.300907 proceedings of DFRWS EU 2020.
Sayakkara, A., Miralles, L., Le-Khac, N.A., Scanlon, M., 2020b. Cutting through the
emissions: feature selection from electromagnetic side-channel data for activity
detection. Forensic Sci. Int.: Digit. Invest. 32 https://doi.org/10.1016/
j.fsidi.2020.300927 proceedings of DFRWS EU 2020.
Soltani, S., Seno, S.A.H., 2017. A survey on digital evidence collection and analysis.
In: 7th International Conference on Computer and Knowledge Engineering
(ICCKE). IEEE, pp. 247e253.
Souvignet, T., Frinken, J., 2013. Differential power analysis as a digital forensic tool.
Forensic Sci. Int. 230 (1-3), 127e136.
Van Eck, W., 1985. Electromagnetic radiation from video display units: an eavesdropping risk? Comput. Secur. 4 (4), 269e286.
Watson, S., Dehghantanha, A., 2016. Digital forensics: the missing piece of the
internet of things promise. Comput. Fraud Secur. 2016 (6), 5e8.
Zankl, A., Seuschek, H., Irazoqui, G., Gulmezoglu, B., 2018. Side-channel attacks in
the internet of things: Threats and challenges. In: Solutions for Cyber-Physical
Systems Ubiquity. IGI Global, pp. 325e357.