IEEE Consumer Electronics Magazine - January/February 2020 - 75

& IN HUMAN-MACHINE INTERACTIONS, CAPTCHA is
commonly used to maintain information security,
it is also a very important and essential part of
the interaction.1 CAPTCHA is used to verify
whether the operator is a machine or a person.
The verification is mainly designed to prevent a
series of illegal machine operations, such as
those carried out by hackers and malicious programs. CAPTCHA is one of the cases of human-
machine recognition, which is necessary to distinguish between people and machines.
Human-machine recognition is carried out in
many different ways. Generally, it can be divided
into two categories: computer-vision-based recognition and operational-behavior-based recognition.
Computer-vision-based recognition (CVBR):
Human behavior and action information are
obtained directly by using sensing and computer
vision technology to carry out analysis and recognition. Such as face recognition, iris detection,2
they are both using computer vision technology
to identify images or video data.
Operational-behavior-based recognition (OBBR):
The related operating records or information is
obtained via user operation on a computer for
analysis and identification. Such as CAPTCHA, it
usually requires user to click or input text or picture to identify whether it is operated by human or
program.3
Regarding CVBR research, Vemulapalli et al.4
recognized human behavior to achieve the man-
machine recognition directly. This article has
been applied in different fields.
Regarding OBBR research,5 Banihashem et al.
proposed a new breakpoint detection technology
for studying computer mouse trajectories called
real-time breakpoint detection. It was used to
identify breakpoints by studying the operation of
a mouse and provides a reference for mouse
operation research.
CAPTCHA belongs to OBBR, i.e., CAPTCHA
code, which includes various forms of code such
as pictures and text. Behavioral CAPTCHA code
is one popular type of CAPTCHA that is primarily
used to verify the operation behavior of a user.
Image CAPTCHA is also widely used, as the operator may be asked to click the specified picture,
and the content of the picture is also extensive.
However, with the development of artificial
intelligence(AI), more and more breakthroughs

January/February 2020

have been made in computer vision, and many
applications for recognition have been developed.6 Image CAPTCHA will eventually be identified by image recognition technology.7 The
generation model can generate an image with
high similarity to the face of user, it is also a threat
to face recognition.8
Both CVBR and OBBR require specific steps to
complete identification or validation in human-
computer interactions, which do not make these
interactions very user-friendly. In addition, the
methods of recognition or verification mentioned
above need to explicitly reveal that verification is
being conducted in an interaction. To some
extent, they remind the operator or intruder to
take precautions.
To summarize the above, the verification
of humans and machines in the common
human-computer interaction has the following
shortcomings:
1. The generation model can produce similar
faces according to the features of the original
face, and the image recognition technology
can also identify the picture verification
code, which can help intruders pass verification. The rapid development of these two
technologies threatens the validity of face
and picture verification.
2. Regardless whether using behavioral, face,
fingerprint or other verification methods,
all these methods require users to perform
related operations, which increases the
length of the interaction process.
3. The above verification is explicit, as these
methods do not offer covert detection.
Explicit verification will remind intruders to
take precautions.
Therefore, in human-machine identification
for man-machine interactions, a better and effective detection method is needed to maintain network security. In actual network security, what
we pay more attention to is the more convenient
human-machine recognition. Computer cursor
tracking analysis has been a popular behavioral
science in recent years.9 Therefore, in this article,
more emphasis has been placed on the human-
computer operation behavior, and the main
research object is the cursor track of a mouse.

IEEE Consumer Electronics Magazine - January/February 2020

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