determination of time of recording with electric network frequency (enf)
DESCRIPTION
Determination of time of recording with Electric Network Frequency (ENF). M.P. Huijbregtse BS, Zeno Geradts PhD AAFS Denver 2009. Outline. Introduction ENF Implications Research Conclusions. Conclusions. Introduction. Implications. Research. Introduction. Electricity network - PowerPoint PPT PresentationTRANSCRIPT
NEDERLANDSFORENSISCHINSTITUUT
Nederlands Forensisch Instituut Laan van Ypenburg 6, 2497 GB Den Haag
Determination of time of recording with Electric Network Frequency (ENF)
M.P. Huijbregtse BS, Zeno Geradts PhDAAFS Denver 2009
NEDERLANDSFORENSISCHINSTITUUT
2/25
Outline
• Introduction ENF • Implications• Research• Conclusions
NEDERLANDSFORENSISCHINSTITUUT
3/25
Introduction
Electricity network• Alternating current • Frequency
• Europe 50 Hz • United States 60 Hz
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
4/25
Introduction
Introduction Implication Research Conclusions
ENF pattern
NEDERLANDSFORENSISCHINSTITUUT
5/25
Introduction – prior research• C. Grigoras, Digital audio recording
analysis – the electric network frequency criterion, International Journal of Speech Language and the Law, vol. 12, no. 1, pp. 63-76 (2005)
• M. Kajstura, A Trawinska, J. Hebestreit, Application of the Electrical Network Frequency (ENF) Criterion – A case of a Digital Recording, Forensic Science International, vol. 155, pp. 165-171 (2005)
• A.J. Cooper, The electric network frequence (ENF) as an aid to authenticating forensic digital audio recordings – an automated approach, Conference paper, AES 33rd International Conference, USA (2008)
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
6/25
Introduction
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
7/25
Introduction
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
8/25
Introduction
• Conditions:• ENF pattern is the same in the whole
network• ENF pattern is “unique”• ENF pattern can be filtered from a
recorder
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
9/25
Implications – build a database
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
10/25
Implications – Build database
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
11/25
Implications Build a database
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
12/25
Software
• Matlab• Java- applet for collecting,
developed by us : www.sourceforge.net enf collector
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
13/25
Software
NEDERLANDSFORENSISCHINSTITUUT
14/25
Implications - filter
• Downsampling naar 300 Hz• Bandpass filteren rond 50 Hz• Spectrogram maken• Piek frequenties bepalen
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
15/25
Implications – Filter signalSpectrogram:
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
16/25
Implications – filter signal
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
17/25
Implications – filter signal
Introduction Implication Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
18/25
Research - Questions
Introduction Implications Research Conclusions
• How unique is an ENF pattern ?• Is it possible to determine the time of
recording ?
NEDERLANDSFORENSISCHINSTITUUT
19/25
Implication
Introductie Praktisch Onderzoek Conclusies
Matching root mean square:
L
nkykxne
L
k
1
2]1[][][
NEDERLANDSFORENSISCHINSTITUUT
20/25
Implication - Uniqueness
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
21/25
Implication - Uniqueness
Length ENF pattern (seconds) Limit (Hz)
60 0.0007
120 0.0015
240 0.0020
420 0.0035
600 0.0040
1800 0.0050
3600 0.0080
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
22/25
Implication best match• Recording of 3600 seconds:
0.053 Hz < 0.080 Hz
• Recording of 600 seconds:0.047 Hz > 0.040 Hz (!)
• Recording of 60 seconds :0.015 Hz > 0.007 Hz (!)
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
23/25
Implication – Best match
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
24/25
Implication – Best match
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
25/25
Second method
yx
L
k
L
ynkyxkxnr
)1(
)]1[)(][(][ 1
Correlationcoefficiënt:
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
26/25
Implication – best match
Introduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
27/25
Conclusions• As in literature:
• ENF pattern is the same in the whole network
• ENF pattern can be extracted from network
• Suggestions:• Matching of short recording (< 600
seconds) false hits are possible• Matching with the correlation coefficent
better than root mean square errorIntroduction Implications Research Conclusions
NEDERLANDSFORENSISCHINSTITUUT
28/25
Future research
• Different recorders• (MP3) compression• Improvement of the current
method