Signal Processing - September 2016 - 149

SP education
Min Wu, Adi Hajj-Ahmad, Matthias Kirchner, Yanpin Ren,
Changshui Zhang, and Patrizio Campisi

Location Signatures That You Don't See
Highlights from the IEEE Signal Processing Cup 2016 Student Competition

Digital Object Identifier 10.1109/MSP.2016.2581738
Date of publication: 2 September 2016

publications such as [3] and [4] and the
nominal value of ENF is 60 Hz in North
references cited therein.
America and 50 Hz in most other parts of
Exploring location information of these
the world. ENF typically does not stay
ENF signals was the topic area of this
constant at its nominal value but rather
year's IEEE Signal Processing Cup (SP
fluctuates around it due to load changes
Cup) competition. The
across the grid and the
"Everyone has gained
SP Cup is an undercontrol mechanisms
graduate competition
knowledge or skills in
that adjust the amount
organized by the IEEE
of power generation
different aspects. The
Signal Processing
to stabilize the freSP Cup is a catalyst."
Society (SPS) in which
quency changes. We
undergraduate students work in teams to
refer to the changing values of the ENF
tackle a real-life signal processing problem.
over time as an ENF signal. An example
Launched in 2014, the SP Cup competition
of an ENF signal can be seen in Figure 1.
has been held annually, and 2016 is the
What makes the ENF particularly relethird edition. To join the SP Cup competivant to multimedia forensics is that audio
tion, undergraduate students are required
or video recordings captured in areas
to form a team. Each team is composed of
where there is electrical activity, such as
one faculty member to advise the team
our offices and living rooms, may capture
members, up to one graduate student to
the ENF variations. In audio, this can be
assist the supervisor in mentoring the
from the ambient power hum, and in
team, and three to ten undergraduate stuvideo, this is due to the near-invisible
dents. Three top teams are selected from
flickering of electrical lighting. As the
the initial round of competition and proexact variations of ENF appear to be ranvided travel grants to participate in the
dom and difficult to predict, the embedfinal competition at the 2016 IEEE Interded ENF traces can serve as an intrinsic
national Conference on Acoustics,
fingerprint that can be used for a number
Speech, and Signal Processing (ICASSP).
of forensic applications. In recent years, it
The final results are shown in "Winners of
has been shown that the embedded ENF
the SP Cup 2016."
traces can be used as a timestamp of
This article provides an overview of the
media recordings as well as an indicator
technical tasks and highlights representative
as to whether or not the media recording
approaches and participants' reflections.
was tampered with [3], [5].
For the 2016 SP Cup, the competition
topic
was shaped around using ENF to
Location signatures
identify
the power grid in which a media
in power and media
recording
was made. Upon examining
As suggested by its name, ENF is the freENF signals collected from different
quency of a power distribution grid. The

IEEE Signal Processing Magazine

hen thinking about using technologies to find one's location, many of
us may consider satellite-based global navigation systems, such as the U.S.
global positioning system (GPS). Smartphones today use GPS type signals and
cellular network signals to determine the
phone user's current location and offer a
broad range of location-related services.
But when there was no such real-time
connection and signal to analyze, for
example, having only a piece of video or
voice recording as an evidence of a news
story or from child exploitation investigations (which have become an important
global issue being addressed by the United
Nation's UNICEF organization), are we
out of luck?
Fortunately, some very weak traces
about the location may be "imprinted" to
become part of the sound or visual signals during the sensing process [1].
These traces cannot easily be seen or
heard by a human, but it is possible to
extract them using signal processing
techniques. The contribution of these
location signatures-perhaps less foreseeable to a layman-come from the variations in electric power supplies [2].
Thanks to information forensics research
in the last decade, the frequency variations, known as electric network frequency (ENF) signals, are being extracted and
exploited to answer a range of forensic
questions. Interested readers may refer to

September 2016

149

Table of Contents for the Digital Edition of Signal Processing - September 2016