IEEE Consumer Electronics Magazine - January/February 2020 - 85

Table 1. Three types of cognitive traits for cognitive identification.
Parameter

Psychological

Activity

Social

Human analogy

What do they
seem like?

How do they act?

With whom do
they connect?

Context

Inherent
aspects

Behavioral
aspects

Social
contexts

Differentiability

Moderate

Strong

Weak

Tenacity

Moderate

Strong

Moderate

Adaptability

Weak

Moderate

Strong

a learnt model. In identification mode, the rating
extraction module retrieves the current rating
data of an unknown user which needs to be classified either as malicious or ordinary user and test it
using learnt model. The learnt model processes
the current and history user's ratings, and produces a unique score value for the unknown user.
This value is used by a decision-making system to
identify that the user is malicious or ordinary.

Figure 1. Cognitive identification framework.

engine compatible with Apache Hadoop that
provides the capability of graph analytics and
machine learning of real-time streaming data. The
machine learning library of Spark helps to extract
the user's psychological and activity traits from
the collected user's data. This module generates
the user's ratings, such as psychological and
activity ratings. The graph computation module
(GraphX) processes collected data related to
user's social connection and produces the user's
social rating. The rating results from spark engine
are passed to the scalable distributed database at
the storage layer, which is further used to the decision-making process whether a user is malicious
or ordinary. Next, the classification layer performs
its task in two modes: learning and identification.
In learning, the rating extraction module retrieves
the history of user's rating data from the distributed database and trains it using intelligent learning algorithm (e.g., strong classifier) that generate

January/February 2020

Cognitive Traits
Cognitive traits play a significant role to
design a cognitive identification framework for
identifying malicious users on SNSs. Selection of
the right traits depends on the following three
prerequisites.
Differentiability: A cognitive trait should provide the ability to differentiate between ordinary
and malicious users.
Tenacity: Tenacity of the cognitive traits also
influences the development of the cognitive
identification framework. This is because most
cognitive traits are regularly extracted from SNS
users, and the interaction between users and
SNSs is of erratic and sporadic patterns.
Adaptability: Since cognitive identification
framework identifies a malicious user based on
his/her psychological and behavioral patterns, a
malicious user can change his/her behavioral
patterns dynamically and try to pose as an ordinary user.
Table 1 defines the three types of cognitive
traits employed in cognitive identification framework for SNSs. Each type is analogous to a class
of human characteristics by which a user on
SNSs can be uniquely identified. For instance,
psychological traits represent basic characteristics of a user's profile on SNSs, on which an

IEEE Consumer Electronics Magazine - January/February 2020

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2020