IEEE Consumer Electronics Magazine - January/February 2022 - 12

Smart Healthcare Security
the section " RAFV: An Overview, " RAFV will allocate
the direction of rotation to the quadrants in
the deepest layer and quadrants in the layers
above will derive their direction from the sum of
their children. For the 128-b key to have a maximum
effect there must be at least 128 quadrants
in the deepest layer to ensure that this key is
represented throughout the rotation pattern. It
is therefore required that a vault must be at least
four layers deep to ensure a minimum of
128 quadrants in the last layer. The number of
quadrants in each layer can be determined using
the following equation:
Quadrantsnumber ¼ 4m1
(2)
where m refers to the layer order. The layers are
ordered starting from one until the deepest layer
of the vault. To compute the total number of
quadrants present within the entire vault we use
the following the equation:
Quadrantstotal ¼ 4m1 þ 4m2 þ ::: þ 4 þ 1:
(3)
Therefore, the fifth layer of a divided vault
(i.e., the result of applying the division four
times starting from the main layer) would contain
256 quadrants and the total number of quadrants
in the entire vault would be 341. It is
important to note that if an adversary had half of
the rotation pattern used by the vault they
would not be able to infer the other half. For
example, if an adversary had managed to acquire
the order of rotation used within a given iteration
of the scheme they would not be able to
determine what direction has been assigned to
each quadrant within the vault.
Furthermore, two types of adversaries may
attempt to compromise the security of the
RAFV-based key agreement. Adversary A is an
adversary that is unaware of any of the locking
elements used by the underlying fuzzy vault.
This means that for every iteration of the rotated
vault they would have to attempt to brute force
the fuzzy vault unlocking process with no guarantee
that they have the original vault. Adversary
B, however, is aware of the elements used
to lock the vault due to their ability to remotely
capture the physiological signals being leaked
12
by the wearer. Therefore, they will be able to
process each vault much quicker because they
will only be concerned with the points that overlap
with the features they have illegitimately
acquired, allowing them to discard chaff points.
However, this does not give the adversary B any
advantage to unlocking the vault as the knowledge
of the locking elements does not provide
any insight into how the vault has been rotated.
PERFORMANCE EVALUATION
To evaluate the performance of RAFV, we will
compare it against the fuzzy vault scheme with
regards to the required execution time. Such a
comparison will enable an accurate assessment
of the additional overhead induced by RAFV.
Both schemes have been implemented in C++17
running on an Intel 8809G (3.10 GHz/4.20 GHz)
processor with a RAM of 2400 MHz C16. Whilst
the performance of such a processor far exceeds
the performance of wireless sensor devices used
in the intended application area (i.e., BAN) this
evaluation aims to provide a like-for-like comparison
between the two schemes. Future work
could, however, look at implementing both
schemes on more appropriate hardware devices.
The evaluation scenario consists of running
both schemes sequentially 100 000 times to measure
their performance for equally sized vaults.
In addition to varying the vault size, from 100 to
5000, to assess its impact on the achieved execution
time for both schemes, the quadrant layers
(and consequently the total number of quadrants)
and the number of applied rotations, relevant
for RAFV only, are also varied to assess how
they affect the execution time and the resulting
search space. More specifically, we evaluated 3
and 4 quadrant layers with 5 different number of
rotations: 16, 32, 64, 128, and 256.
Figure 4(a) shows the time taken to construct
an RAFV over many parameters: various sizes,
number of rotations in addition to the time taken
to construct an identically sized fuzzy vault. This
figure demonstrates that the execution time for
either the fuzzy vault or RAFV scheme increases
in line with the vault size. With regards to RAFV,
its execution time increases further as more
rotations are carried out, this is expected as
more quadrants are being selected for rotation
IEEE Consumer Electronics Magazine
IEEE Consumer Electronics Magazine - January/February 2022

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2022
