Most models of earthquakes attempt to reproduce the observed scaling laws o
f seismic events: the Gutenberg-Richter frequency magnitude distribution, b
ut not the Omori law for aftershocks and the multifractal distribution of h
ypocenters location. Many of these models are based on the idea of Self-Org
anized Criticality (SOC). These are dynamic systems which organize themselv
es into a transitional state and can reproduce the Gutenberg-Richter distri
bution, but generally do not reproduce the space-time distribution. Here, w
e suggest a model based on a fractal geometry: the two sides of a fault are
modeled by means of a fractal surface. As a first step, one of them is sli
pped of a random amount with periodic boundary conditions, then new contact
points between the surfaces are found. The area surrounded by these points
is assumed to be proportional to the area of the earthquake. The size dist
ribution of events is in good agreement with the observed Gutenberg-Richter
law and the local fluctuations of the b value are explained in terms of va
riations of the fractal dimension of the surface. Also the multifractal dis
tribution of earthquakes in space is well-reproduced with global properties
not depending on the fractal dimension of the surface. However, we are not
able to obtain something similar to the Omori law simply because we do not
control the time evolution of the model.