Periodic account activity and automated money laundering detection
| Date | 01 October 2004 |
| Pages | 295-297 |
| DOI | https://doi.org/10.1108/13685200410810001 |
| Published date | 01 October 2004 |
| Author | Carl Young |
| Subject Matter | Accounting & finance |
ANALYSIS
Periodic Account Activity and Automated Money
Laundering Detection
Carl Young
INTRODUCTION
The automated detection of money laundering can
involve assumptions about `normal' movement of
assets in and out of accounts. Money laundering is a
companion crime, where individuals seek to legitimise
assets derived from explicitly illegal activities. A var-
iety of factors have been cited by ®nancial industry
regulators pursuant to identifying money laundering
activity. Speci®cally, know-your-customer, customer
due diligence, monitoring client activity, and geo-
graphic considerations have been important com-
ponents of a comprehensive anti-money laundering
programme.
Recently, regulators have also cited the importance
of analysing account activity relative to client peer
groups as a means of identifying unusual behaviour.
In order for transactional behaviour to be deemed
unusual and therefore suspicious, transaction patterns
must be seen to deviate from past behaviour, peers'
behaviour, or exceeding pre-de®ned thresholds. The
process of identifying risk in this way is therefore
inherently probabilistic, and detecting a money laun-
dering `smoking gun' is non-trivial.
If it is possible, however, to make certain assump-
tions about transactional behaviour, either based on
an individual's past record or through information
gleaned during the account opening process, various
analytic methods may be applied to help ®nancial
institutions identify unusual behaviour and associated
risk. Once identi®ed, these risk factors would be based
on deviations from anticipated account behaviour,
and methods can be used to ®lter `noise' resulting
from a preponderance of legitimate transactions.
The method of detection proposed herein is quite
simple and well known in engineering/physics appli-
cations. It is, however, believed to be unique in its
application to the anti-money laundering detection
problem. An important next step would be to validate
its eectiveness against real data.
Speci®cally, the proposal here is to analyse an
account history as a time series of asset movements
that can be analytically characterised by a function,
f(t). The Fourier transform of f(t) is then computed,
yielding a function of frequency, F(w). This quantity
can be used to compute the so-called Power Spectrum,
thereby specifying the `power' or magnitude of asset
movement in the account as a function of frequency.
Standard digital ®ltering techniques would then be
applied to designated frequency components, based
on a presumption of risk-relevance for speci®ed fre-
quencies or account behaviour types. This technique
lends itself to the application of digital ®ltering tech-
niques, with the goal of easily identifying transactions
that deviate from an established pattern, and then
applying further account review if warranted.
PERIODIC ACCOUNT BEHAVIOUR
Account histories can often be characterised by a series
of deposits and withdrawals of assets as a function of
time. These can be viewed as a progression of peaks
and troughs, similar to waves appearing in the physical
world. Depending on its intended use, an account
history for an individual re¯ects a variety of regular
deposit/withdrawal phenomena, such as deposits asso-
ciated with salary pay cheques, withdrawals corre-
sponding to mortgage payments, loan repayments
etc. Although a one-time transaction or a number of
unrelated transactions might not be uncommon for a
given account, each may be cause for concern if it
deviates from anticipated behaviour.
Representing an account history as a time series, and
applying Fourier analysis to the function so created
would yield frequency information with respect to
that account. Speci®cally, a Fourier representation of
account transactions yields detailed spectral features
of a transaction history that would immediately
distinguish between regular and isolated account
behaviour.
Page 295
Journal of Money Laundering Control Ð Vol. 7 No. 4
Journalof Money Laundering Control
Vol.7, No. 4, 2004, pp. 295± 297
#HenryStewart Publications
ISSN1368-5201
To continue reading
Request your trial