IEEE Technology and Society Magazine - September 2015 - 45

Brain-computer interfaces are a fast-growing class of
non-fictional exocortices. A large variety of BCIs are currently being proposed and developed for various applications ranging from medical to commercial, including
advertising, marketing, and gaming. For example, the
Nielsen Company acquired NeuroFocus in 2008, "for the
development of neural engineering technologies aimed at
better understanding customer needs and preferences"
[6]. In May 2013, Samsung, in collaboration with the University of Texas, demonstrated how BCIs could be used to
control mobile devices [10]. In the same month, the first
neurogaming conference was held [7].
Several consumer-grade BCIs and software development
kits, developed by, for example, Emotiv Systems [5] and
NeuroSky [8] currently exist on the market. In addition, the
same companies have recently introduced the concept of
app stores for BCIs, in order to facilitate expansion of BCI
applications (as described in [15]). Future BCIs will likely be
simpler to use and will require less training time and user
effort, while enabling faster and more accurate decoding of
users' intentions. It is easy to see that Wiener was correct
when he predicted that information technology will give rise
to prostheses for able-bodied people which will "give them
powers that humans never had before" [16], [43].
To paraphrase Wiener's concerns, this progress poses
not only new possibilities for the future, but raises questions about users' privacy. In the recent security literature, researchers introduced several BCI-enabled
malicious applications, referring to them as brain spyware [23], [34]. These applications can be used to
extract users' private information, such as credit card
PINs, from recorded neural signals.
As BCIs become widespread, more sophisticated
spying applications become easy to imagine. Leveraging
recent results in neuroscience, it is possible to extract
private information about users' memories, prejudices,
religious and political beliefs, as well as about their possible neurophysiological disorders [15]. The extracted
information could be used for marketing purposes, or
to manipulate or coerce users, or otherwise harm them.
The impact of BCI malware could be severe in terms of
privacy and other important values.
This work is inspired by Wiener's Information Ethics
methodology, which can be summarized by the following steps [16]:
1) Identify an ethical problem or positive opportunity
regarding the integration of technology into society.
2) If possible, apply existing policies, using precedent
and traditional interpretations to resolve the problem
or to benefit from the opportunity.
3) If existing policies appear to be ambiguous or vague
when applied to the new problem or opportunity,
clarify ambiguities and vagueness.

september 2015

∕

4) If precedent and existing interpretations, including
the new clarifications, are insufficient to resolve the
problem or to benefit from the opportunity, revise the
old policies or create new ones.
5) Apply the new or revised policies to resolve the problem or to benefit from the opportunity.
We take a Value Sensitive Design approach to understanding and designing privacy-preserving brain-computer technology. We start from the hypothesis that the
time to identify and address privacy and security challenges arising from uses of exocortices is now.

Preliminaries
The initial development of BCIs was motivated by the
needs of individuals having certain physical disabilities,
and the recognition of the potential benefits these systems might offer. A variety of BCI platforms exist (and
are being developed) for assistance, augmentation, and
repair of cognitive and sensorimotor capabilities in conditions such as spinal cord injuries and amyotrophic lateral
sclerosis. In addition, BCIs are increasingly being used
in non-medical applications, such as gaming, entertainment, and marketing. For example, researchers have proposed a method of predicting success of online games by
analyzing a user's electromyographic (EMG) signals (electrical signals produced by a user's skeletal muscles) over
the first 45 minutes of the game [17].

Components of a BCI
From an engineering perspective, a BCI is a communication system between the brain and the external environment, consisting of inputs (user's neural activity), outputs
(external world commands), and components - translating inputs to outputs, referred to as signal acquisition
and signal processing. For an image of a high-level block
diagram of a typical BCI, please refer to [15, fig. 1] in the
June issue of this magazine.
In this system, communication messages between the
brain and the environment are typically encoded as electrophysiological signals [45], [46]. Based on the type of communication signal (and recording location), BCIs can range
from completely implanted devices, where electrode arrays
are surgically implanted into the brain; to partially invasive,
such as electrocorticography (ECoG), where electrodes are
typically implanted inside the skull and on top of the dura;
to non-invasive devices. Most non-invasive BCIs are based
on electroencephalography (EEG), but other biosignals,
such as EMG or electrooculograms (EOG) can also be used
to control devices. While known to be susceptible to noise
and signal distortion, EEG signals are easy to measure. In
addition, EEG-based BCIs have relatively low cost and risk,
which makes them the most widely used BCI devices [44].
In this article, we focus on non-invasive BCIs.

IEEE TEchnology and SocIETy MagazInE

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - September 2015