A Prospect Theoretic Approach for Trust Management in IoT Networks under
Manipulation Attacks
Abstract
As Internet of Things (IoT) and Cyber-Physical systems become more
ubiquitous in our daily lives, it necessitates the capability to measure
the trustworthiness of the aggregate data from such systems to make fair
decisions. However, the interpretation of trustworthiness is contextual
and varies according to the risk tolerance attitude of the concerned
application. In addition, there exist varying levels of uncertainty
associated with an evidence upon which a trust model is built. Hence,
the data integrity scoring mechanisms require some provisions to adapt
to different risk attitudes and uncertainties.
In this paper, we propose a prospect theoretic framework for data
integrity scoring that quantifies the trustworthiness of the collected
data from IoT devices in the presence of adversaries who try to
manipulate the data. In our proposed method, we consider an imperfect
anomaly monitoring mechanism that tracks the transmitted data from each
device and classifies the outcome (trustworthiness
of data) as not compromised, compromised, or undecided. These outcomes
are conceptualized as a multinomial hypothesis of a Bayesian inference
model with three parameters. These parameters are then used for
calculating a utility value via prospect theory to evaluate
the reliability of the aggregate data at an IoT hub. In addition, to
take into account different risk attitudes, we propose two types of
fusion rule at IoT hub– optimistic and conservative.
Furthermore, we put forward asymmetric weighted moving average (AWMA)
scheme to measure the trustworthiness of aggregate data in presence of
On-Off attacks. The proposed framework is validated using extensive
simulation experiments for both uniform and On-Off attacks. We show how
trust scores vary under a variety of system factors like attack
magnitude and inaccurate detection. In addition, we measure the
trustworthiness of the aggregate data using the well-known expected
utility theory and compare the results
with that obtained by prospect theory. The simulation results reveal
that prospect theory quantifies trustworthiness of the aggregate data
better than expected utility theory.