|
|
  |
 |
|
|
|
|
|
September 28, 2009 Technological Deception
All of us have witnessed the wonder of movie
special effects. Some may have even considered that our system is
now capable of deceiving the masses if they desire. Other
technologies are just as persuasive. The convergence of digital cameras and
computer graphics programs have made photography an area of special
concern now for investigators. Analog photography in the past was
difficult to tamper with. Now, suspicions that digital images can
be convincingly altered have forced investigators to look for some
unalterable means to gather photographic evidence. To this end,
they have reverted to using Polaroid cameras because the entire process
is done at once by one mechanism.1
Images
taken by film cameras requiring a separate outside processing are
thought to also be easily altered by digital software. All of
this is done to protect the truth and to prevent the dismissal of cases
on grounds of untrusted evidence. Some may have come to suspect that audio
also can be altered. When I worked at Nortel, I tested a product
called DRAMS (digital recorded announcement messages.) It is my
understanding that messages were spliced together from words and
phrases. The following citation from Nortel technical
documentation confirms this belief (italics are mine): DRAM MECHANICS The basic unit
of DRAM speech data is the
phrase. A phrase corresponds to an English phrase, sentence, or
group of sentences. A given phrase may be a single word, an
entire sentence or even a group of sentences stitched together at the
operating company’s discretion. An announcement
trunk is made up of a set
of members. A call that is routed to a specific announcement has
one of the trunk members chosen by the CC for its terminator.
Depending on operating company requirements, up to 255 subscribers may
be simultaneously connected to a single channel. An earlier version of a similar document
says: Phrases are connected together by the DRA to form different messages.3 Although memory technology now can contain
messages in their entirety, this has not always been so. Nortel
introduced their line of fully digital telephone switches in 1976, a
time when memory capacities would not support long recordings of
audio. The first personal computers only had about 16KB of RAM,
in time increasing to 64KB. The documentation that I had readily
at hand, dated 1988, lists circuit packs with 128KB and 256KB memory
capacity, giving a theoretical capacity of 16 and 32 seconds of audio,
respectively. I remember testing older circuit packs with less
memory than this perhaps as low as 8KB or 16KB. A 16KB memory
module would only hold 2 seconds of audio (8-bit telephone audio is
sampled at 8000 samples per second). The need to conserve memory
by splicing messages together from words, syllables, or short phrases
is evident. I remember hearing operator recordings
during this period of time; that they were stilted and would not pass
for real speech. Much has happened since then. In the
summer of 1994, as a new scapegoat, a coworker was quizzing me about
words with body language that betrayed some discomfort as of
deceit. When they left the work station, because I knew of the
technology to splice words, I spoke to whoever might listen, “Are you
trying to gather words to put in my mouth?” Soon afterward, my
attempt to call my supervisor was interrupted by a recorded
message. The message was spliced together from words from
different persons. It was done so skillfully, that if it had been
the words of one person it would have passed for unaltered
speech. Surely this was an affirmative answer to my previous
query, even if only pertaining to their technical capability. You will ask, how did their skills
improve? On many lines. First, they did advanced research
in voice technology as evidenced by a Unix-based voice recognition
product that I worked, called Network Application Vehicle. The
kinds of skills and knowledge gained in this research (and that of
others) would be applied to the use of highly capable visual audio
editors in the task of effectively splicing the audio. These
tools (even Audacity or Nero wave editor) allow one to see whether the
joined clips align properly. The software’s filters could be used
according to user’s specialized knowledge to smooth out any
anomalies. A store of syllables and words from different sentence
contexts would ensure that the flow and accenting of the resulting
audio would be convincing. Certainly, their skills improved for
very concrete reasons after all and are perhaps not ones that novices
would be expected to accomplish. Covert psychic technology could be added to this list, but is beyond the scope of this article.4 Beware!
1Alan
Axelrod
and Guy Antinozzi, 2003, “The Complete Idiot's
Guide To Criminal Investigation”, page 261 Document History |
|