Database Reference
In-Depth Information
distance constraint and the interval constraint. Such fundamental event
constructors can be combined to form complex RFID events.
For example, a company uses RFID tags to identify asset items and
employees in the building, and only authorized users (superusers) can
move the asset items out of the building. When an unauthorized em-
ployee or a criminal takes a laptop (with an embedded RFID tag) out
of the building, the system will send an alert to the security personnel
for response. Such complex RFID event pattern (
Q
2
) can be expressed
in the RCEDA language as follows:
WITHIN(E1
∧¬
E2, 5sec)
Here events
E1
and
E2
are two primitive events:
E1 = observation('r2', o1,t1)
,
type(o1)= 'laptop'
E2 = observation('r2', o2, t2)
,
type(o2) = 'superuser'
.
Based on the event specification described above, RFID rules can be
defined to support data filtering, data transformation, data aggregation
and real-time monitoring. The RFID rule for event
Q
2
is shown below:
DEFINE E4 = observation('r4', o4, t4), type(o4) = 'laptop'
DEFINE E5 = observation('r5', o5, t5), type(o5) 'superuser'
CREATE RULE r5, asset monitoring rule
ON WITHIN(E4
∧¬
E5, 5sec)
IF true
DO send alarm
Here
r5
and
asset monitoring rule
are unique rule id and rule name
respectively.
WITHIN
(E4
E5, 5sec)
is the event part of the rule.
send alarm
is an action to be performed while the specified event occurs.
According to the defined rule, an alert alarm will be issued when an
unauthorized employee takes a laptop out of the building.
∧¬
3.3 RFID Complex Event Detection Models
While RFID event specification languages provide an expressive way
to specify complex RFID events, the detection of such events is much
more challenging. The detection models in active databases have lim-
itations on supporting RFID events. Automata-based model [77] and
the PetriNet-based model [78, 79] require that all the timestamps of the
constituted events are in total order. Tree-based model [54] and graph-
based model [80] does not support time constraints. All these traditional
models can not be directly used for RFID complex event detection.
An NFA-based complex RFID event detection model supplemented
with Partitioned Active Instance Stacks (PAIS) [42] is proposed to sup-
port complex RFID event detection, especially for event backtracking
