Content uploaded by Stefan K. Braun
Author content
All content in this area was uploaded by Stefan K. Braun on Jul 31, 2014
Content may be subject to copyright.
Forensic Evidence of Copyright Infringement by Digital Audio Sampling
Analysis - Identification – Marking
Stefan K. Braun
Faculty of Management. Comenius University
PhD Student, Comenius University
Bratislava, Slovak Republic
mail@stefanbraun.eu
ABSTRACT
In recent years, the number of attempts to use digi-
tal audio and video evidence in litigation in civil
and criminal proceedings has increased. Technical
progress makes editing and changing music, film
and picture recordings much easier, faster and bet-
ter. The methods of digital sampling differ from the
conventional pirated copy in that using a sample
involves extensive changes and editing of the origi-
nal work. Different digital sampling methods make
the technical analysis and the legal classification
more difficult. Targeted analysis methods can clear-
ly identify a case of sampling and belong to the
main field of forensic analysis. If persuasive evi-
dence of an unauthorized use of sampling cannot be
produced, the proof is useless in the legal process.
Labelling technologies that are applied correctly
make an important contribution to the effective
detection of unauthorised sound sampling. There are
hardly any holistic approaches that integrate the
problem of sound sampling into the fields of analy-
sis, identification, and labelling. In combination
with specific technical protective mechanisms
against sampling, an unauthorised use of samples
protected by copyrights can be prevented or re-
duced. Using and sampling somebody else’s piece
of music or video can be a copyright infringement.
The copyright and the neighbouring rights of per-
forming artists and the neighbouring rights of pho-
nogram producers are affected by the consequences
of illegal sampling. Part 1 of the article introduces
the problems of digital audio sampling, Part 2 de-
scribes the typical manifestations of sampling, Part
3 illustrates various analytical procedures for the
detection of audio sampling and Part 4 shows the
identification by labelling strategies.
KEYWORDS
audio · authentication · bootlegging · digital tech-
niques · single sound sampling · ENF · Electric
Network Frequency · forensics · forensic audiology
· real-time frequency analysis · cryptography ·
neighbouring rights · melody · mash-up · mix pro-
duction · multi-sampling · phase inversion · remix ·
sample medley · sound sampling · sound separation
· spectrogram · spectrometer measurement · sound
collage · sound sequence sampling · copyright ·
watermarking
1 INTRODUCTION
1.1 The Problems and Classification of
Digital Sound-Sampling
The word “sample” in this context stems from
the piece of equipment known as a “sampler”.
The sampler is supplied with sound information
by integrating sound or microphone recordings.
From the fed-in oscillation curves, samples are
taken and stored. With the use of modern soft-
ware, removed samples can be, for example,
transposed in pitch and tempo, mutilated, trans-
formed, tampered or mixed as desired [1, 2].
From the sample source voices, instruments,
rhythms and parts of melody can be removed
(“sampled out”) and incorporated into a new
production. The purpose of sampling is the
simple and inexpensive way of adopting desired
sounds, instruments, or voices without having
to invest in studio production costs, time and
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
187
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
effort. Furthermore, the sound characteristics of
performers can be imitated and used as inspira-
tion without their knowledge or consent.
Users of samplers not only utilize notes but also
sound from a specific production. The ar-
rangement of individual sounds and timbres can
be created, on the one hand, in the studio and,
on the other, directly on the digital recording
computer [1]. “Sound”, “timbre” and “tone” are
used more or less synonymously in literature.
The limiting factor is that from a physical point
of view, timbre is only one of the many compo-
nents of sound [3].
Sounds and melodies can be generally adopted
from both existing music productions and re-
cordings. In contrast to this, there are sound
databases that can be downloaded from the
internet and also physical data carriers such as
sound libraries.
In addition to shorter sound excerpts of a few
bars or seconds, smaller melody parts, the so-
called “licks” and smaller sequences are sam-
pled. A specific sampled music sample there-
fore includes also the generated sound [4]. If
there are, in addition to a certain sound, enough
of these samples available to the user, he can
put these together like a “mosaic” to create a
“new” work. A very common form of sampling
is taking foreign compositions from actual re-
cordings into new music and film productions.
Often pitches and characteristics are changed to
differing degrees when adopting single tones or
tone sequences in the sampling process.
Processing. The processing of a musical work
is always associated with a transformation.
When composing, the melodic, harmonic and
rhythmic form is changed. When this is text, it
is reworked, modified, supplemented, replaced
completely or translated into another language,
for example. The result of such a major rear-
rangement is a newly created work. The cover
version shows the necessary individuality in the
form of intellectual and approval-requiring
creation [5]. The prerequisite is that the trans-
formation in turn has the appropriate quality of
“work”. It should be determined which musical
design elements cause the creative peculiarity
of the work. To be considered in this context in
particular are the tonal system, the duration of
the tone, timbre, volume, rhythm and melody.
Processing eligible for protection. Processed
work which is eligible for protection requires a
recognizable creative performance of the editor,
so that resulting from the compositional change
or expansion of the musical substance of the
original, a new, independent work is created. In
contrast to such works which are eligible for
protection are those which use an original work
and take the musical substance of the original
essentially unchanged and transfer the musical
text of the original faithfully (e.g. editorial ser-
vices) [6]. Works that have been created using
other works or foreign melodies must be
marked with the appropriate copyright informa-
tion. For free works no permission for process-
ing has to be sought from the originator. Pro-
tected works require this permission. Process-
ing is the key feature when considering whether
the original is eligible for protection [5]. It is
crucial that the new work distinguishes itself
from the old one and not only repeats an al-
ready existing one; the aesthetic overall impres-
sion of the new piece must not be present in the
original work [7].
Melody. The melody is, in occidental music,
the most important parameter and main infor-
mation carrier. Together with the harmony it is
the most important forming structure in music.
The term melody includes three elements:
Harmony (harmonizing of tones), rhythmos
[sic] (temporal structure) and logos (text).
Melodies are differentiated in their function and
their classification as a vocal melody (range,
phrase length) or instrumental melody [8]. The
melody forms a self-contained tone system
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
188
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
(characteristic). It retains its own character even
when accompaniment (rhythm) is eliminated or
the sounds replaced (transposed). In music for
easy listening and pop music, the vocal parts of
the melody are considered to be the characteris-
tic that can be assigned to the relevant song.
1.2 Services not Eligible for Protection
Typical techniques and thus ineligible for pro-
tection include mere conversions of sentences
or sentence parts of a multi-part musical work,
slight changes of melody, harmony and rhythm,
or individual noise elements if the basic charac-
ter of the original work remains the same [9].
Certain, recurring basic repeats or patterns,
such as chord sequences, classic song structures
or common elements of music are not eligible
for protection [10]. Insignificant tonal varia-
tions, slight shortening or extensions taking into
account the compositional or textual original
work are permitted in this context [10]. Excep-
tions are to be seen under certain circumstances
with regards to fingering in music course books
when this characteristic forms the tone. The
transposition of the pitch of the original is also
one of the criteria ineligible for protection and
does not change the melody.
Criteria for Activities not Eligible for Protec-
tion.
─ Lack of originality.
─ Insignificant, minor changes.
─ Use of an original work, borrowing of partial
works.
─ Transposition to a different key or pitch for
technical artistic reasons.
─ Instrumentation and timbre of individual in-
struments, merely replacing an instrument.
─ Adaptation of the melody to the vocal abili-
ties of the singer.
─ Making changes to the rhythm, replacement
with another standard rhythm.
─ Note-for-note transcription of existing voices
to another instrument.
─ Supplementing of performance indication,
elaboration, fingering, applying punctuation.
─ Addition, change of phrasing.
─ Tempo and volume adjustments.
─ Doubling of voices.
─ Addition of accompanying voices in parallel
motion (e.g. in the third or sixth).
─ Reduction of existing parts in the score of a
piano movement.
─ Editorial services (publication of a pre-
existing musical work).
─ Digitization or compression into an MP3 file,
for example.
2 TYPICAL MANIFESTATIONS OF
SAMPLING
Cooper [11] divides audio editing into three
levels: 1) Editing / tampering on a basic level,
directly in the original material, during or after
the recording; 2) Editing / tampering on an in-
termediate level, containing several fields cop-
ied from one or more original sources for a new
recording; 3) Editing / tampering at a high level
by means of appropriate editing and sound
processing software. The edited version will
then function as a “new original”.
According to their type of use, the sampling
techniques can be divided into single-tone sam-
pling and melody sampling. Single-tone sam-
pling distinguishes again between the actual
sampling of a single-tone and a variant called
“Multi-Sampling”, one of the economically
most important and technically difficult to de-
tect sampling forms. It is referred to colloqui-
ally as “sound sampling”. The parties involved
in each sampling are always the originator or
author, the performing artist and, in the case of
indirect sampling, the record producer. If a
digital sample is used, there is inevitably al-
ways a reproduction of works or parts of works.
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
189
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
2.1 Origin of the Sound Material
Sampling of the Artists' Own Sound Mate-
rial. Sound material can be recorded by the
artists themselves or recorded and then sam-
pled. This is usually done where there are cer-
tain fragments repeated in a musical work.
Sampling is also carried out when certain fig-
ures of a piece have a repetitive character and
do not differ in dynamics, articulation and
rhythm. With this approach, difficult figures
and phrases have to be recorded only once [12].
Sampling of Foreign Sound Material. Much
more sampling material comes from external
sources [12] such as sound recordings or indi-
vidual tracks from multi-track tapes. Further-
more, so-called “factory sounds” and sound
archives exist, for example, on CD or as
downloads from internet archives.
Natural Sounds. These are divided into signals
produced by oneself and others as well as natu-
ral sounds, meaning sounds not shaped by hu-
mans including animal sounds, machinery and
everyday sounds [13] and meteorological
noises [12].
2.2 Single-Tone Sampling
Direct Single-Tone Sampling. Under direct
single tone sampling, sampling of individual
instrumental sounds is understood. Here, a cer-
tain characteristic sound, for example, an in-
strument, a voice or a sound is taken in isola-
tion, digitized, fragmented, and then imported
into the sampling computer [12]. Using the
keys of keyboards, the sound can be allotted to
a button and then played. If there are sounds in
different pitches, volumes and articulations,
music can be played and modelled with specific
musical characteristics. This process provides
unrestricted access to the original sound of a
music production.
Indirect Single-Tone Sampling. Single-tone
indirect sampling is the term used to refer to the
acquisition of sampled sounds from existing
recordings, mostly audio recordings. A single
tone can thus be isolated and the obtained
sound then processed. The acquisition of single
tones from a ready-mixed multi-track produc-
tion by frequency superpositions of the single-
tones and instrumental tracks later mixed to-
gether is not quite so simple. A single tone from
single tracks of a recording, however, is very
easy to remove and include and of high quality
[1].
Multi-Sampling. The term multi-sampling is
used when several individual notes with differ-
ent pitch intervals and volumes are distributed
on a sampler keyboard. The distribution usually
takes place according to the original pitch. Of-
ten tones of mixed productions are extracted
which have superimposed frequencies of other
instruments. If only one sound as in the single-
tone sampling is extracted, this would have to
be transposed to a different pitch, which would
lead to frequency distortions in any existing
secondary frequencies. Therefore, different
sounds according to their pitch ranges are ex-
tracted from different points of a piece in order
to avoid this negative effect. An additional op-
timization is achieved by the blending (posi-
tional crossfading) of the samples with each
other [1].
2.3 Melody Sampling
Contrary to the sound use of the single tone
sampling, tone sequences sampling is about the
(partial) adoption of melodies, harmonies and
rhythms and the subsequent collage-like com-
position of new musical works. In general, a
sequence of sampled parts from well-known
music productions is used to maintain the rec-
ognition effect [12]. A variety of procedures
can be distinguished.
Mixed Productions (Sample Medley). In
mixed production consecutive characteristic
music parts of a few seconds or bars are sam-
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
190
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
pled and successively linked together in a
newly created mixed production. Here, the new
mixed production either contains parts of sam-
ples [1] or, in extreme cases, consists entirely of
such. By using adjustment of the tempo the
individual samples must be adapted, where
necessary, before the mixing takes place. The
purpose of this approach is the recognition ef-
fect of the sampled work parts. The more
clearly the recognition of parts of the origina-
tor’s work, the more successful the goal of the
mixed production was implemented. Such
mixed productions are created in the pop and
dance genres by disc jockeys. Such productions
were used before digital sampling technology
existed, carried out by hand and the much more
complicated and time-consuming tape cutting.
Sound Collages. Unlike mixed productions,
sound collages disguise their origin [14]. In-
stead of stringing together sound samples, into
sound collages these are layered over each
other (“batch processing”). It is not unusual for
several layers of samples to be superimposed.
For example, a melody sequence can be taken
as a sample from work 1, a rhythm from work 2
and a guitar sequence from work 3. In general,
the individual samples must then be adjusted
with regards to volume, tempo, pitch and tim-
bre, so that they fit together in a new produc-
tion, often cut as a “loop”. As with mixed pro-
ductions, the sound collages may consist either
in part or entirely of samples.
Cover Versions. The sampling technique with
cover versions and remixes is understood as
“hit-recycling”. Either the whole work or parts
thereof, for example, the refrain, are taken from
the original and backed with new rhythms and
sounds. The purpose is the audible sound adap-
tation to new listening habits. Cover versions
(interpretations of an earlier original) can be
made without using the sampling technique.
The sampling technique is still used con-
sciously and for economic reasons, however, to
maintain the successful part of the original. As
with the mixed productions, sampled parts
should be recognized [1].
If the artist leaves the limited scope for inter-
pretation set for cover versions and moves to-
wards a processing with independent creative
input into the piece, this change is subject to
approval.
Remixes. The remix follows the same rules as
processing. Successful hits are frequently re-
released as a remix. Individual tracks of a
multi-track tape are often completely “broken
down into pieces” and recomposed and remixed
along with new recordings. There are also
mixed sound effects, new recordings of instru-
ments and a far-reaching change in the sound of
the material. The remix, however, can take
place with the extraction of a sample [12].
Mash-Up. Mash-ups (also known as bootleg-
ging, bastard pop or collage) have been enjoy-
ing increasing popularity for years. At the be-
ginning of the 1990s, it was usually only 2 dif-
ferent pop songs whose vocal and instrument
tracks were mixed with each other to form a
remix [12]; today there are multi-mash-ups with
several dozen mixed and sampled songs, artists,
video sequences and effects. It is a challenge to
mix this combination of different styles to new
danceable tracks.
The mash-up is a mix of sound collage and
mixed productions. Usually known sequences
of two or more (multi-mash-up) existing works
are mixed to create a “new” work. The samples
used are layered over each other (sound col-
lage), as well as in series (mixed production).
The incorporation of large parts of the original
in the mash-up is the rule. In sampling, how-
ever, it is rather the exception [12].
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
191
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
3 ANALYSIS METHODS FOR
DETERMINING THE USE OF A
SAMPLE
Evidence of sampled parts in a musical work
can be achieved by means of different methods
of analysis.
3.1 Musical Aspects
Under certain circumstances, a simple listening
test is sufficient. As a rule, a direct comparison
of the musical notation is carried out. Since
most samples were changed in speed and pitch,
it can be helpful, to adapt these in terms of
pitch and tempo to the original before starting
the analysis.
Pitch changes and temporal extension have
qualitative limits if a realistic overall impres-
sion should remain. Deviations of about 15-
20% produce audible noise and alienate the
original. This can be desirable for creative rea-
sons. Often sampled parts are superimposed
with other instrument and vocal tracks. A sim-
ple separation is then no longer possible.
3.2 Physicals Aspects
Analysing and measurement methods provide
evidence of the use of sampling.
Electric Network Frequency Analysis (ENF).
With regard to the validation of digital audio
and video recording a common method recog-
nized by expert forensics is the Electric Net-
work Frequency Analysis (ENF). Each mains
power supply leaves a characteristic frequency,
a so-called "mains hum". This may not be audi-
ble but the oscillations can be detected in an
audio file [15]. If digital recording devices are
used such as cameras or audio recorders to rec-
ord voice, music or film recordings these, in
addition to the actual content, store the network
frequencies of 50 or 60 Hz. This happens with
battery powered devices in the same way [16].
The frequency of the carrier never has exactly
the same value. The random fluctuations in the
power supply are the result of the differences
between produced and consumed current. The
actual ENF signal can be extracted by using
band-pass filters that filter out, for example, in
a 50 Hz power supply the range 49-51 Hz. In-
terruptions or irregularities in the phase re-
sponse can be an indication of tampering. The
network frequency behaves in effect as a tem-
porary digital watermark. The effectiveness of
this method, however, is dependent on such a
network signal existing at all [15]. In most situ-
ations, a visual comparison of the spectrogram
with the frequencies which are stored in an
ENF database is sufficient. More detailed stud-
ies require measurement and analysis in certain
short time slots, which are compared to each
other. With this method it is even possible to
determine the exact location and the exact time
of production of a recording. Corresponding
reference samples prepared by continuous re-
cording of the network frequencies in power
networks (such as the German or European
electricity grids) are a prerequisite [16].
Microphones also leave a particular frequency
spectrum in the audio material. Should several
different spectra show up in a recording, this
can also be an indication of tampering [15]. The
evaluation of the digital audio recording by
detecting the exact measurements, the compari-
son, and a mapping of the individual frequen-
cies in the reference database is, therefore, of
great importance.
Suitable methods are the spectrogram represen-
tation, “re-sampling”, real-time-frequency
analysis (spectrometer measurement) and the
phase inversion.
Besides ENF analysis the spectrogram repre-
sentation, “re-sampling”, real-time frequency
analysis (spectrometer measurement) as well as
phase inversion are appropriate forensic meth-
ods.
Spectrogram Representation. In a spectro-
gram display the spectral density of a signal is
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
192
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
displayed over time. Figure 1 shows in A the
recording A (an original music recording from
1990) and B the recording B (unauthorized ed-
iting by sampling removed from the original A
and mingling with new instrumental tracks in
2007). The recordings were first equalized in
tempo and pitch before the examination by
means of a spectrogram representation and then
directly compared (see “A” (left channel) as
recording A and “B” (right channel) recording
as B). A time section of 6 seconds is depicted.
This corresponds to about 4 synchronous paral-
lel bars from the two recordings. The Y-axis
shows the frequency spectrum of 0 Hz (Hertz)
to about 20 kHz and depicts approximately the
hearing range of the human ear1. In the lower
area, the lower frequencies are shown and the
higher frequencies at the top. The horizontal X-
axis is the time axis. Frequency is the number
of periods that are run through in a second. The
unit of frequency is the Hertz (Hz). An oscilla-
tion is composed of a positive and a negative
half-wave, i.e. the to-and-fro swing of the elec-
trons is called an oscillation, wave or period
[17].
With this representation, the audio material can
be visualized. The representation in the fre-
quency spectrum is used to gain direct access
both to specific frequency ranges as well as
certain time ranges in contrast to standard
waveform processing (see Figure 1, Area C)
which is always performed for the entire fre-
quency domain. These frequency ranges can be
shown in colour by means of analysis software.
High and low frequencies are represented by
different colours. The intensity and the level of
the frequencies are displayed in a colour spec-
trum that extends from blue and white (the
highest intensity) to purple and black (the low-
est intensity). In simple terms, a bell sound in a
piece of music can, for example, be reduced,
replaced or removed by using the “Copy &
Paste” software function to copy a part without
1 The hearing range (auditory sensation area) of human
ear is from about 16-21 Hz to 16-19 KHz.
a bell and insert it over the desired place. In
spectral processing, there are diverse modes
that can be used. For example, it is possible to
reduce levels by means of band, low and high
pass filters (“damping”) - the peak level is
blurred by mixing the frequencies and thus they
“disappear” or are covered up. Furthermore, it
is possible to transform the dynamics without
changing the actual frequency content (“disper-
sion”).
A
B
C
Figure 1. Comparison of the recordings A and B in the
spectrogram representation (Source: Stefan K. Braun).
In Figure 1 a bell can be clearly seen in the
sampled recordings A and B (red arrow). It lies
in the frequency range of approximately 5400
Hz. In the AB-comparison it is very clear to see
that all the different frequencies correspond, the
patterns of both images are identical. The fre-
quency spectrum of recording B is much richer.
This is due to the mixed instrumental tracks
added to the recording. The temporal distribu-
tion of the frequency phase and the significant
characteristic features such as the existing bell
in the recording have not been changed by the
sampling. The area C represents the waveform
processing. Visual procedures such as the spec-
trogram representation are important methods
for aiding detection of manipulation.
“Re-Sampling”. Under certain circumstances,
a sampling procedure can be carried out via a
so-called “re-sampling”. Here, in simple terms,
the numerical values of the digital samples are
compared with those of the original. This pre-
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
193
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
supposes, however, that there are identical
comparative pieces. Usually the samplings used
do not exist in isolation, but in the final product
mixed together inseparably with other audio
and instrument tracks, distorted with effects and
changes in tempo and pitch. A direct compari-
son is no longer possible.
Spectrometer Measuring. In sampling, a digi-
tal copying process cannot always be compared
purely by listening. With spectrometer meas-
urement a coherent frequency diagram can be
displayed and a very accurate and detailed real-
time frequency analysis performed. In this case,
the frequency spectrum is represented as a lin-
ear graph. Peak levels are depicted as short
horizontal lines showing the last reached
maximum values (see Figure 2). Spectrometer
measurements are also used in forensic analy-
ses, e.g. vocal comparisons in the field of
criminology.
Figure 2. A frequency spectrum in a real-time frequency
analysis with a linear graph at a randomly selected point
in time of the investigated sample (Source: Stefan K.
Braun).
In Figures 2 and 3, the amplitude of the wave of
+ / -0 dB (decibels) is represented by -96 dB on
the vertical Y-axis, the horizontal X axis indi-
cates the frequency band of 0 Hz to about 16
kHz. As described above, the recordings A and
B are compared directly. In the authenticity
analysis, the determination of the originality
and continuity of the recordings and the detec-
tion of changes are of particular importance
[18]. For real-time frequency analysis a random
location of the samples to be examined was
selected and fixed as a linear graph. The graphs
show the result of a very similar, almost match-
ing curve. Their frequency forms correspond in
the typical manifestations in the characteristic
points (e.g. shallow rise, steep climb, strong
peaks between 500 Hz and 2700 Hz, falling
from 7500 Hz). Within the investigated samples
of 6 seconds duration all investigated linear
graphs of the frequency spectrum show a rela-
tively similar curve in terms of characteristics
and patterns. Figures 4 and 5 show two or more
overlapping linear graphs. The relatively simi-
lar curves from randomly selected positions in
time on the same samples show clear similari-
ties between the original A and sample pro-
cessed B.
A
A
B
B
Figure 3. A frequency spectrum in a real-time frequency
analysis with a linear graph at a different point in time of
the same sample (Source: Stefan K. Braun).
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
194
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
Figure 4. A frequency spectrum in a real-time frequency
analysis with two overlapping linear graphs at a random-
ly selected point in time of the same sample (Source:
Stefan K. Braun).
Figure 5. A frequency spectrum in a real-time frequency
analysis with three overlapping linear graphs at a ran-
domly selected point in time of the same sample (Source:
Stefan K. Braun).
The problem may be verification when a sam-
pling was not created by copying, but by an
extensive technical sound remake. Here there is
a difference in the technical and legal view.
While in terms of law, a remake “sample” can
still be considered as such, it is technically a
different object. If a sample is taken from an
original, it can be determined relatively easily
due to whether the frequency plot of the linear
graphs is the same or different in the analysed
sample. For example, physical characteristics of
the same or different audio tracks of vocals can
be represented by this method. Adopted or re-
made instrument passages can be revealed and
checked for sameness with this method. Even
non-audible differences of different blowing
techniques for brass instruments or different
striking techniques with keyboard instruments
can be seen in the graph representation [13]. It
is not possible to achieve congruent sound and
frequency structures by imitating ways of play-
ing and singing. If they are identical, everything
points to a sampled adoption of the original.
The limits of an identical representation of the
linear graphs are reached when the samples in
one object which are being compared are
changed dramatically with respect to sound and
are superimposed with other vocal and instru-
ment tracks.
Phase Inversion. In recording studio technol-
ogy, phase reversal (phase inversion) is often
used to correct wrongly polarized audio signals
in the phase. In order to achieve certain effects,
phases with correct polarity can also be re-
versed deliberately. Using this, undesired and
reverse-poled phases can be added/mixed with
the phases of the original signal, so that they
cancel each other out, in whole or in part. For
example, in a piece of music with vocals, the
vocals are “filtered out” by phase inversion in
order to obtain an instrumental or karaoke ver-
sion.
In the forensic evidence of phase reversal, a
destructive interference is sought; the matching
points (oscillations) of the samples cancel each
other out. An oscillation is composed of a posi-
tive and a negative half-wave, and thus corre-
sponds to a full circle of 360 degrees [17]. If
two sine-phases in the fundamental frequency
are shifted 180 degrees of the phase, they are
opposed (mirrored or inverted) and so cancel
each other out completely.
If two or more waves are added, their ampli-
tudes are reinforced; this is referred to as con-
structive interference. If the waves cancel each
other out, destructive (complete) interference is
the term used.
Theoretically, both recordings must be com-
pletely identical in this experimental arrange-
A
B
A
B
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
195
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
ment, i.e. tempo, pitch, volume and the course
of the wave form match completely. If in a re-
cording, a phase inversion is performed and this
phase is mixed together with the other identical
recording without phase inversion, it results in a
complete cancellation of the part concerned.
In a study of the phase reversal, destructive
interference was sought in order to mutually
cancel the corresponding parts of the samples.
Under practical conditions, the physical align-
ment of both recordings on exactly the same
pitch is very difficult. The more accurate this
process is, the greater the cancellation in the
end. In the next step, recording A is inverted in
phase and levelled with the pitch of recording
B. Then both phases are superimposed. The
result is shown in Figure 6. While the phases do
not cancel themselves out completely, they
clearly correlate with each other. This correla-
tion is particularly evident in the direct compar-
ison with the unprocessed recording B.
Comparison objects are seldom completely
identical in practice. A phase cancellation is
therefore mostly only partially possible. The
affected sample part has partial cancellations.
What can be heard after a partial phase inver-
sion is a clear “flanging effect”. This effect is
caused by artificial zeroes which are the result
of the cancellation of the audio signal in the
frequency spectrum. At the same time, in the
previous phase reversal, a phase shift will take
place, which causes a shift duration (“delay”).
Now both the (partially) erased places and also
the shifting of the phases to each other are au-
dible
“Flanging” altered audio signals produce a kind
of “floating” effect. Often the effect is de-
scribed like a jet (“jet effect”) which moves
through the music [19]. In simplified terms, the
“flanging” effect is similar to that of a tape and
tape recorder. If a spool is “braked” by hand,
then it accelerates again when released. This
creates the effect of “flanging”.
Complete cancellation cannot be achieved even
with perfect alignment of pitch as the examined
images A and B are different in their character-
istics. This is mainly due to, as mentioned
above, the superposed instrumental and rhythm
tracks in recording B.
A
B
P
Figure 6. Phase inversion. A (input A, phase inverted), B
(B recording, normal phase), P (mixed phases from the
recordings A and B and audible “flanging”). Recording B
is “shorter” at the parts with the included sampling at the
end of the sample, i.e. it stops earlier than at the corre-
sponding part in recording A. Due to this, the flanging
effect stops at the end of this part. At this point the previ-
ous partial cancellation is particularly apparent (Source:
Stefan K. Braun).
4 IDENTIFICATION BY LABELLING
STRATEGIES
There is almost no effective protection that pre-
vents unauthorized copying. In the last 20 years
or so, the affected industries have developed
and used the most diverse digital copy protec-
tion and labelling systems. Known systems
include Digital Rights Management (DRM), the
Content Scrambling System (CSS), different
types of holograms, signatures such as RIFD,
Serial Copy Management System (SCMS) or,
for example, digital watermarks. For novice
users there might be restrictions in use as not all
the playing devices are able to deal with the
copy protection mechanisms such as the DRM
restrictions. The technically versed professional
is, regardless of the legal regulations, capable
of getting round these precautions more or less
easily. Although overall markings such as
holograms, bar codes [20] or ISRC codes (In-
ternational Standard Recording Code) [21]
identify the product (recorded music, digital
file) in terms of its originality, they do not pro-
tect or prevent a possible further illegal use. Of
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
196
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
importance is a modular approach between the
requirements of sound sampling, in conjunction
with a proper identification method: Protection
and recognition of very small clippings that are
superimposed with other signals in foreign pro-
ductions reappear. All procedures which can be
used have a main problem in common: the
more they cost, the less value these are in a
practical use. Fundamentally it must be distin-
guished between “data hiding” and “water-
marking”. While data hiding conceals informa-
tion in the medium or in a channel, the water-
marking binds the information into the medium.
Data hiding is used interchangeably with "in-
formation hiding", although the latter is more
likely to be used for the cryptic method [22].
The following procedures seem appropriate for
marking, identification and authentication of
sound samples for further use:
4.1 Cryptographic Processes
Cryptographic processes can be divided into
asymmetric, symmetric and hybrid2, as well as
strong and weak methods.
According to Lynch / Lundquist a cryptic-
secure data exchange is confronted with the
following system requirements: Identification,
authentication, verification, non-repudiation
and privacy. If all five demands are met, this is
referred to as a secure data exchange [23].
Asymmetric, cryptic processes are character-
ized by the fact that digital signatures have a
private and a public code. With the use of the
private codes it is ensured that only the owner
of the product rights can assign an individual
signature [20]. A test of the encryption is pro-
vided by the public code. Signatures, e.g. in the
form of identification numbers (“Identification
Keys”) in connection with a verification data-
base allow the tracking of marked objects
(“Tracking & Tracing”).
2 https://www.datenschutz.rlp.de/downloads/oh/ak_oh_
kryptographie_version1.pdf
4.2 “Watermarking”
The watermarking technology is a promising
technology for the protection and prosecution
of copyright infringement. The basic technique
and main focus of research in digital water-
marking consists of an integral, invisible [22]
“interweaving” of identification (copyright in-
formation, names, logos, etc.) with the main
channel without interfering with or impairing
this. Audio signals (music and speech), images,
movies, software, e-books and texts can be pro-
vided with individual markings in this way
[24].
There are two important main groups of water-
marking use: 1 Piracy resistant use, which pre-
vents an attack on the watermark. Applications
are copy-protection measures, “fingerprint”
techniques and other preventive measures (e.g.
hash functions). 2 A use that is weak in terms
of being piracy resistant, the watermark is dis-
solved or minimally changed in the case of a
piracy attack. When the watermark has been
changed or is absent, copies of the originals are
no longer recognized as originals [25].
There are important requirements for the label-
ling: 1 The easy readability of the watermark in
retrospect, 2 Resistance to destruction, 3 The
receipt of the signal in the case of the use of
very small excerpts of the original file [25] and
4 The additional information must not be per-
ceptible to the human ear [22].
There are several, often conflicting, properties
that are the focus of watermarking: The amount
of hidden or inserted information, the robust-
ness and security of the data, the invisibility
and the reading of the introduced data [22].
Labelling and identification systems which are
based on an authentication and so distinguish
the copy from the original can be used inde-
pendently or with a database [20]. A check on
the authenticity of the watermark and the con-
trol of the authentication is done, for example,
using database systems. For audio files, for
example, a watermark can be set as an “inaudi-
ble” frequency over the actual audio frequency
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
197
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
band. To read the information it needs the same
algorithm, a “Watermark Key”, which was nec-
essary for the earlier interweaving of the infor-
mation. The recognition of copyright infringe-
ment takes place via a verification comparison
on the database server. Disadvantages of such
systems are a not quite closed security chain, as
markers are not created directly at the premises
of the copyright owner, but in the sales shop. If
only digital files using the watermark process
are detected, a direct use of recorded music
media and trade on exchange platforms cannot
be prevented. Piracy resistance has limits with
the use of watermarking technology. A frequent
copying and transforming creates a “fuzzy”,
unreadable watermark. A significant advantage
is in the aforementioned limitations of the pres-
ervation of the watermark even with format
changes, compression, filtering, re-sampling,
re-quantization, as well as recognizing the vio-
lation of even the smallest excerpts, as they
occur with the sound sampling [24].
5 Conclusion
In principle, only the adoption of free or law-
fully licensed works is allowed for processing
as a sample. If it is unclear whether sampling
should be carried out, a sample-clearing with
their respective rights holders and collecting
societies can help.
With regard to the validation of digital audio
and video recording a common method recog-
nized by expert forensics is the Electric Net-
work Frequency Analysis (ENF). With this
method it is even possible to determine the ex-
act location and the exact time of the produc-
tion of a recording. Visual procedures such as
the spectrogram representation are important
methods for aiding detection of manipulation.
In the authenticity analysis, the determination
of the originality and continuity of the re-
cordings and the detection of changes are of
particular importance.
Sound Sampling will continue to win in impor-
tance and new extraction methods (sound sepa-
ration) which can extract the whole melody will
exacerbate the problem of piracy. On the other
hand, the improved analysis and marking proc-
esses such as the watermarking technology of-
fer more possibilities for the protection and
detection as well as prosecution of copyright
violations.
References
[1] M. Häuser, Sound und sampling: Der Schutz der
Urheber, ausübenden Künstler und Tonträgerher-
steller gegen digitales Soundsampling nach deut-
schem und US-amerikanischem Recht. Disserta-
tion. München: Beck, 2002.
[2] E. Adeney, “The sampling and remix dilemma:
What is the role of moral rights in the encourage-
ment and regulation of derivative creativity,”
(English), Deakin Law Review, vol. 17, no. 2, pp.
335–348, 2012.
[3] T. M. Jörger, Das Plagiat in der Popularmusik.
Dissertation, 1st ed. Baden-Baden: Nomos Ver-
lagsgesellschaft, 1992.
[4] B. Wessling, Der zivilrechtliche Schutz gegen
digitales Sound-Sampling: Zum Schutz gegen
Übernahme kleinster musikalischer Einheiten
nach Urheber-, Leistungsschutz-, Wettbewerbs-
und allgemeinem Persönlichkeitsrecht. Dissertati-
on. Baden-Baden: Nomos Verl.- Ges, 1995.
[5] M. Pendzich, Von der Coverversion zum Hit-
Recycling: Historische, ökonomische und rechtli-
che Aspekte eines zentralen Phänomens der Pop-
und Rockmusik. Dissertation. Münster: LIT, 2004.
[6] Ohne Verfasser, GEMA: Schutzfähige Bearbei-
tungen freier Werke. Journal Article
[7] U. Loewenheim and B. von Becker, Handbuch
des Urheberrechts. Textbook, 2nd ed. München:
Beck, 2010.
[8] R. Amon, Lexikon der Harmonielehre: Nach-
schlagewerk zur durmolltonalen Harmonik mit
Analysechiffren für Funktionen, Stufen und Jazz-
Akkorde. Textbook. Wien, Stuttgart: Doblinger;
Metzler, 2005.
[9] R. Moser, Handbuch der Musikwirtschaft. Text-
book, 6th ed. Starnberg u.a: Keller, 2003.
[10] G. Berndorff, B. Berndorff, and K. Eigler, Musik-
recht: Die häufigsten Fragen des Musikgeschäfts ;
die Antworten. Textbook, 6th ed. Bergkirchen:
PPV-Medien, 2010.
[11] A. J. Cooper, Detection of copies of digital audio
recordings for forensic purposes. Milton Keynes:
Open University, 2006.
[12] P. Wegener, Sound Sampling: Der Schutz von
Werk- und Darbietungsteilen der Musik nach
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
198
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
schweizerischem Urheberrechtsgesetz. Dissertati-
on. Basel: Helbing Lichtenhahn, 2007.
[13] R. Münker, Urheberrechtliche Zustimmungser-
fordernisse beim Digital Sampling. Dissertation.
Frankfurt am Main, New York: P. Lang, 1995.
[14] T. Meschede, Der Schutz digitaler Musik- und
Filmwerke vor privater Vervielfältigung nach den
zwei Gesetzen zur Regelung des Urheberrechts in
der Informationsgesellschaft. Dissertation. Frank-
furt/Main, New York: P. Lang, 2007.
[15] R. Korycki, “Time and spectral analysis methods
with machine learning for the authentication of
digital audio recordings,” (Englisch), Forensic
Science International, vol. 230, no. 1-3, pp. 117–
126, 2013.
[16] C. Grigoras, “Applications of ENF criterion in
forensic audio, video, computer and telecommu-
nication analysis,” (Englisch), Forensic Science
International, vol. 167, no. 2-3, pp. 136–145,
2007.
[17] H. Meister, Elektronik: Mit Versuchsanleitungen
und Rechenbeispielen, 8th ed. Würzburg: Vogel-
Buchverlag, 1986.
[18] B. E. Koenig and D. S. Lacey, “An Inconclusive
Digital Audio Authenticity Examination: A
Unique Case,” (Englisch), Journal of Forensic
Sciences, vol. 57, no. 1, pp. 239–245, 2012.
[19] J. Webers, Tonstudiotechnik: Handbuch der
Schallaufnahme und -wiedergabe bei Rundfunk,
Fernsehen, Film und Schallplatte, 4th ed. Mün-
chen: Franzis, 1985.
[20] M. Abramovici, Kennzeichnungstechnologien zum
wirksamen Schutz gegen Produktpiraterie: Mit
Ergebnissen aus Projekten MobilAuthent, O-Pur,
EZ-Pharm. Research Project. Frankfurt am Main:
VDMA, 2010.
[21] M. Schäfer and T. Hansen, ISCR - International
Standard Recording Code: Das ISRC-Handbuch.
Textbook. Available:
http://www.musikindustrie.de/fileadmin/news/pub
likationen/vb_isrc_handbuch.pdf (2014, Feb. 17).
[22] M. Faundez-Zanuy, J. J. Lucena-Molina, and M.
Hagmüller, “Speech Watermarking: An Approach
for the Forensic Analysis of Digital Telephonic
Recordings*,” (Englisch), Journal of Forensic
Sciences, vol. 55, no. 4, pp. 1080–1087, 2010.
[23] D. C. Lynch and L. Lundquist, Digital money:
The new era of Internet commerce. Journal Arti-
cle. New York: Wiley, 1996.
[24] S. V. Dhavale, “Lossless Audio Watermarking
Based on the Alpha Statistic Modulation,”
(Englisch), IJMA, vol. 4, no. 4, pp. 109–119,
2012.
[25] M. A. Nematollahi and S. A. R. Al-Haddad, “An
overview of digital speech watermarking,”
(Englisch), Int J Speech Technol, vol. 16, no. 4,
pp. 471–488, 2013.
International Journal of Cyber-Security and Digital Forensics (IJCSDF) 3(3): 187-199
199
The Society of Digital Information and Wireless Communications, 2014 (ISSN: 2305-0012)
