Can anyone give me a lead for this;

As the audio comes in to drive, the stream itself needs to be encrypted to
be made tamper-proof, not just read-only. If anyone does as much as change a
meta value or cut a sample, it must ring alarm bells. (It's a
security/evidence thing)

I'm not getting far on google with this, have you been here before?

--
dE|_

In article , dE|_
wrote:
Can anyone give me a lead for this;

As the audio comes in to drive, the stream itself needs to be encrypted
to be made tamper-proof, not just read-only. If anyone does as much as
change a meta value or cut a sample, it must ring alarm bells. (It's a
security/evidence thing)

You don't say anything about what kind of 'audio' you mean, or what kind of
system the 'drive' is in. That might help.

Nor do you say why what you need could not be done by watermarking the
audio externally to the digital recording process. e.g. by low level ENF.

'As it comes in' would have to be better defined in terms of method. You
could encyrpt on a bit-by-bit or sample-by-sample level against a
max-length sequence or similar. Or use blocks encrypted, etc. The best
choice would depend on what you want and how the system is to be used.

If using a secure system like Unix/Linux then the 'admin' or the specified
user would have to 'know' the encryption keys or equivalent, or at least
have them via shadow, so you can never absolutely guarantee that evidence
has not been tampered with. All you can do is give some level of protection
and detection. You also don't say if those using the machine to record
'audio' should also have the ability to recover the data. i.e. decrypt it.

In forensic terms 'tamper proof' is in practice a relative term, not and
absolute, and also depends on physical chain of evidence handling and the
persons involved operaing in a suitable, documented, manner.

I'm not getting far on google with this, have you been here before?

Are you a member of the AES? if so you can look at the recent papers in
JAES and conferences about 'forensic audio'. They will give you some ideas
on this, and some individuals you could contact who might advise.
Presumably if you are doing this for professional purposes you would also
be paying for any help. :-)

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

"Jim Lesurf" wrote:
Can anyone give me a lead for this;

As the audio comes in to drive, the stream itself needs to be encrypted
to be made tamper-proof, not just read-only. If anyone does as much as
change a meta value or cut a sample, it must ring alarm bells. (It's a
security/evidence thing)

You don't say anything about what kind of 'audio' you mean, or what kind
of
system the 'drive' is in. That might help.

Nor do you say why what you need could not be done by watermarking the
audio externally to the digital recording process. e.g. by low level ENF.

Sorry, it does not exist yet and the options would be all mine which is why
I have not specified. Due to the amount there would be, mp3 on hard disk
would be most practical.

I had thought about watermarking with a slow flanger. What is this ENF you
speak of?

'As it comes in' would have to be better defined in terms of method. You
could encyrpt on a bit-by-bit or sample-by-sample level against a
max-length sequence or similar. Or use blocks encrypted, etc. The best
choice would depend on what you want and how the system is to be used.

I'll look into that, sounds interesting.

If using a secure system like Unix/Linux then the 'admin' or the specified
user would have to 'know' the encryption keys or equivalent, or at least
have them via shadow, so you can never absolutely guarantee that evidence
has not been tampered with. All you can do is give some level of
protection
and detection. You also don't say if those using the machine to record
'audio' should also have the ability to recover the data. i.e. decrypt it.

In forensic terms 'tamper proof' is in practice a relative term, not and
absolute, and also depends on physical chain of evidence handling and the
persons involved operaing in a suitable, documented, manner.

I'm not getting far on google with this, have you been here before?

Are you a member of the AES? if so you can look at the recent papers in
JAES and conferences about 'forensic audio'. They will give you some ideas
on this, and some individuals you could contact who might advise.
Presumably if you are doing this for professional purposes you would also
be paying for any help. :-)

There is a chance I could be doing this for professional purposes in the
future, I'm looking into all the basic options first.

--
dE|_

In article , dE|_
wrote:

"Jim Lesurf" wrote:
Can anyone give me a lead for this;

Nor do you say why what you need could not be done by watermarking the
audio externally to the digital recording process. e.g. by low level
ENF.

Sorry, it does not exist yet and the options would be all mine which is
why I have not specified. Due to the amount there would be, mp3 on hard
disk would be most practical.

I suspect you will find that forensic workers would be suspect of lossy
formats like mp3 as they do not represent a full record of the source.
However I can't say more as I don't know enough about the specifics.

They may also want the orginal HD to examine, not just files copied from.
This lets them check the context actually on the disc itself to see if
there are any signs of tampering external to the file payload.

Similarly, they might end up requiring source code for any routines used to
process the recordngs. So open source systems have an advantage here in
evidential terms even if you then have less cover for any aspects you want
to keep commericial in confidence. However I am biassed here. As an
(ex-)academic I regard open source as an inherently more satisfactory
approach to any process which might need to be scrutinised or checked.

I had thought about watermarking with a slow flanger. What is this ENF
you speak of?

Electric Network Frequency. In effect, allowing the sound recording to
have a small amount of 'mains hum'. (Which tends to happen anyway, even
when you *don't* want it. ;- )

The frequency of mains hum varies with time, and differs from time to time
and with the location (i.e. which mains network covers your area). Hence
it can and has been used as evicence that recordings were recorded when
and where claimed and have no breaks or edits - or not! Bit like an audio
forensic version of dating a tree by its growth rings. Surprisngly, this
technique has been used a number of times in court, etc, and apparently
works well. Even for recordings made on battery systems where the mics
picked up the hum as interference.


On 31 Oct in uk.rec.audio, dE|_ wrote:

"Jim Lesurf" asked: snip
Are you a member of the AES? if so you can look at the recent papers
in JAES and conferences about 'forensic audio'. They will give you
some ideas on this, and some individuals you could contact who might
advise.

No, I'm not. I'll Google 'forensic audio' next,

You may find names like Catalin Grigoras, or maybe Koenig and Lacey of BEK
TEK LLC may help. But there may well be others who are more relevant to
what you require. Afraid I can't say for sure as this isn't really my
field. I've only learned about this via casual interest in Journal papers,
etc.[1]

if you've got any pdfs of these papers you could pass on to help with my
job challenge just swap spambusters with del.

There are a number of them on the AES website and/or on CDROMs they
supply. Afraid I don't specialise in Forensic Audio so don't have (or know
the contents of) the bulk of them. Hence if you are serious about this I
would recommend membership. Or finding if a local Uni library has JAES at
least. There are also, I think, Journals on Forensic Science that will deal
with such issues. Again a Uni library may help.

If your recordings would be likely end up as court evidence then there is a
great deal involved in establishing that the methods you choose would be
accepted as evidence. Otherwise 'legal eagles' will find all kinds of
excuses for them to be ruled out of consideration. So I'm afraid this isn't
simply a matter of decent engineering and knowing about crypography or
being able to watermark/sign data.

However the levels you'd need to go to would depend on the nature of the
confidence required and the sort of contest the results would be involved
in deciding. Internal arguments within a company could be decided as the
company chose.

Slainte,

Jim

[1] The usual process where you start reading journals and books for one
purpose, only to find you've diverted onto reading something totally
different which seemed interesting when encountered. :-)

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

"Jim Lesurf" replied to:
Can anyone give me a lead for this;

Nor do you say why what you need could not be done by watermarking the
audio externally to the digital recording process. e.g. by low level
ENF.

Sorry, it does not exist yet and the options would be all mine which is
why I have not specified. Due to the amount there would be, mp3 on hard
disk would be most practical.

I suspect you will find that forensic workers would be suspect of lossy
formats like mp3 as they do not represent a full record of the source.
However I can't say more as I don't know enough about the specifics.

They may also want the orginal HD to examine, not just files copied from.
This lets them check the context actually on the disc itself to see if
there are any signs of tampering external to the file payload.
snip

Scrap the mp3 idea, found this baby and the manufacturer is about 5 miles
from the company who wants the secure recording system.
http://www.cedaraudio.com/products/c...cambridge.html

Also rather than constant recording to mp3 I want to get an auto trigger
system on this so the uncompressed recording kicks off when a signal comes
down the line.

--
dE|_

"Jim Lesurf"

Electric Network Frequency. In effect, allowing the sound recording to
have a small amount of 'mains hum'. (Which tends to happen anyway, even
when you *don't* want it. ;- )

The frequency of mains hum varies with time,


** It wanders a small amount, like 0.1%, around the nominal frequency.


and differs from time to time and with the location (i.e. which mains
network covers your area).

** Mains networks cover huge geographical areas, sometimes whole countries -
within which the frequency is locked in phase.


Hence
it can and has been used as evicence that recordings were recorded when
and where claimed and have no breaks or edits - or not! Bit like an audio
forensic version of dating a tree by its growth rings. Surprisngly, this
technique has been used a number of times in court, etc, and apparently
works well. Even for recordings made on battery systems where the mics
picked up the hum as interference.


** Sure - you can say if it was in a 50Hz or a 60 Hz country.

That really narrows things down.



...... Phil

"Phil Allison" wrote in message
...

"Jim Lesurf"

Electric Network Frequency. In effect, allowing the sound recording to
have a small amount of 'mains hum'. (Which tends to happen anyway, even
when you *don't* want it. ;- )

The frequency of mains hum varies with time,


** It wanders a small amount, like 0.1%, around the nominal frequency.

The tolerance within the EU is +/- 1%. It's unlikely to be much less
elsewhere.


and differs from time to time and with the location (i.e. which mains
network covers your area).

** Mains networks cover huge geographical areas, sometimes whole
countries - within which the frequency is locked in phase.

Even if it only tells you which country it was recorded in, that may well be
significant.

Hence
it can and has been used as evicence that recordings were recorded when
and where claimed and have no breaks or edits - or not! Bit like an audio
forensic version of dating a tree by its growth rings. Surprisngly, this
technique has been used a number of times in court, etc, and apparently
works well. Even for recordings made on battery systems where the mics
picked up the hum as interference.


** Sure - you can say if it was in a 50Hz or a 60 Hz country.

The difference between 50Hz and 50.5Hz (or even 50.05Hz) is easy to detect
and measure these days. Discontinuities in the 50Hz waveform, caused by
editing, can also be readily detected.



David.

In article , Phil Allison
wrote:

"Jim Lesurf"

Electric Network Frequency. In effect, allowing the sound recording to
have a small amount of 'mains hum'. (Which tends to happen anyway,
even when you *don't* want it. ;- )

The frequency of mains hum varies with time,


** It wanders a small amount, like 0.1%, around the nominal frequency.

Yes. The significant point being that if you look in court papers, and
journal papers for the AES and Forensic Science you can see that the
variations with time show patterns that can be reliably used to tell the
date and time recordings were made since many of the details change
unpredictably from day to day.

and differs from time to time and with the location (i.e. which mains
network covers your area).

** Mains networks cover huge geographical areas, sometimes whole
countries - within which the frequency is locked in phase.

Indeed. And in Europe sometimes cover a series of linked countries.

That is one of the useful features as it means Forensic bodies can monitor
and collect data on the patterns for later comparison with evidential
recordings. As has happened in a number of cases already reported in the
journals.

AES members will have seen papers on this recently. The September 2009
issue of JAES discusses the points I make above. Contains a couple of
papers and a feature article on Audio Forensics. Must admit I was surprised
to read just how useful the technique had turned out to be!

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

Jim Lesurf writes:

snip

Slainte,

Jim

[1] The usual process where you start reading journals and books for one
purpose, only to find you've diverted onto reading something totally
different which seemed interesting when encountered. :-)

Which is how I came to follow this thread :-)

And the finally I got my act together and looked up "Slainte" and found:
http://www.awa.dk/glosary/slainte.htm

Cheers,

/Martin

"Jim Lesurf" asked:
snip
Are you a member of the AES? if so you can look at the recent papers in
JAES and conferences about 'forensic audio'. They will give you some ideas
on this, and some individuals you could contact who might advise.

No, I'm not. I'll Google 'forensic audio' next, if you've got any pdfs of
these papers you could pass on to help with my job challenge just swap
spambusters with del.

Cheers,

--
dE|_