Blockchain-Enabled Security in Electric Vehicles Cloud and Edge Computing

Hong Liu, Yan Zhang, and Tao Yang

Centerless Trust: There is no center

EVCE computing is an attractive network para- digm node in peer-to-peer communications,
involving seamless connections among het- in which data exchange is performed
erogeneous vehicular contexts. It will be a trend without pre-assigned trust
along with EVs becoming popular in V2X. The EVs act relationships.
as potential resource infrastructures refer- ring to Collaborative Intelligence: An EV
both information and energy interactions, and there makes a lim-
are serious security challenges for such hybrid cloud ited contribution to specific
and edge computing. Context-aware vehicular processing; the coor- dinated EVs will
applications are identified according to the jointly address problem solving for
perspectives of information and energy inter- actions. crowd intelligence.
Blockchain-inspired data coins and energy coins are Spatio-Temporal Sensitivity: An EV’s
proposed based on distributed consen- sus, in which exchanged information and energy
data contribution frequency and ener- gy contribution are sensitive data with obvious
amount are applied to achieve the proof of work. spatio-temporal attributes for the
Security solutions are presented for securing data provenance requirement.
vehicular interactions in EVCE computing. Unlike traditional system
Introduction infrastructures, the EVCE confronts
Electric vehicles cloud and edge (EVCE) comput- ing serious security challenges due to the
is an attractive network paradigm involving seamless legal entities and attackers being
connections in heterogeneous vehicular contexts to equipotent participants owning equal
aggregate distributed electric vehi- cles (EVs) into a privilege. For instance, an EV
common resource pool, and to invoke the EVs for provides dynamic traffic information
locally flexible usage [1]. In EVCE computing, and idle energy as distributed
information flow and energy flow are dynamically resources. An attacker could receive
exchanged during the vehi- cle-to-anything omni-bearing messages via open
communications, including vehicle- to-grid (V2G), interfaces and wireless
vehicle-to-infrastructure (V2I), and vehicle-to-vehicle communication channels. In order to
(V2V), to achieve collaborative data sensing, address the security issues,
information analyzing, and energy sharing. Due to the blockchain is introduced as a
data sensitivity and context complexity, vehicular potential solution with two main
applications confront seri- ous security issues. features. Decentralization means that
The emergence of autonomous vehicles the data accounting, storage,
promotes the popularity of the combination of maintenance, and trans- mission are
distributed vehicular resources. EVs constitute a performed based on distributed com-
noteworthy portion of connected services with puting capabilities, and there is no
energy, communication, and computation resources core node for centralized
to perform processing at the network edge. management; Co-participation means
Meanwhile, EVs’ idle resources may be aggregated to that all the participants will join in
establish a mobile resource pool for collaborative block transac- tions based on
utilization [2]. Thus, a vehicular ecosystem will be consensus algorithms.
established based on the dis- tributed and Toward the blockchain, cryptographic
aggregated EVs through their activity cycles, in which algorithms are applied to establish
information and energy interac- tions are achieved mutual or multi-party trust
among EVs, sensors, roadside units (RSUs), and local relationships. Mass collaboration is
aggregators (LAGs) for interactive applications. performed by collective self-interests,
The coexistence of hybrid cloud computing and edge and the possible data uncertainty is
computing is becoming the trend of future vehicular regarded as a mar- ginal element.
applications, which have the fol- lowing three Any block records and stores a
characteristics. receipt to a link with the previous
block, and a new block is attached to
the ledger only if the corresponding messages pass achieve enhanced security protection.
authentication by the majority of participants [3]. This The remainder of this work is
special data structure provides better robustness organized as
under a single point of failure, and avoids tampering follows. The next section introduces
attack with data traceability. Currently, block- chain- the net- work architecture and
based key management [4], anonymous multi- context-aware vehicular applications.
signatures [5], and secure software defined network Following that, we present security
architecture [6] have been proposed and leveraged to requirements and distributed
enhance security. In this article, the blockchain is consensus. Then we
applied during information and energy interactions to
Digital Object Identifier: Hong Liu is with East China Normal University and Shanghai Trusted Industrial Control Platform Co., Ltd.; Yan Zhang (corresponding
10.1109/MNET.2018.170 author) is with University of Oslo; Tao Yang is with The Third Research Institute of the Ministry of Public
0344 Security.

78 0890-8044/18/$25.00 © 2018 IEEE

IEEE Network • May/June
2018
present the security solutions for both road environment detection. The sensing data is gathered by EVs
information and energy interactions. The for satisfying diverse functional requirements. Much vehicular
final section draws a conclusion. data temporarily exists or entirely remains on the EVs, while some
may be transmitted to and from other entities.
NETWORK ARCHITECTURE The EVs are distributed computing nodes to aggregate data
AND CONTEXT-AWARE together, and the associated data is applied for cooperative
services, including traf- fic query, driving assistance, and
APPLICATIONs NETWORK entertainment sharing. The EVs realize resource reallocation to
satisfy specific individual demands. Here, informa- tion interactions
ARCHITECTURE are generally based on V2I and V2V communications. For
Figure 1 illustrates the network instance, an EV com-
architecture of EVCE computing, in which
there are two comput- ing modes for
extending mobile cloud infrastruc- tures
to the EVs as the network edges.
EV Cloud Computing: The EVs’ idle energy,
computation, and communication
resources can be aggregated into a
common pool. The EVs are regarded as
mobile cloudlets based on vehicular ad
hoc networks (VANETs) and other
networks. The connected EVs are
gathered for providing cooperative
services while moving round different
areas or spending hours in parking lots.
The cloud computing mode highlights
extending tradition- al cloud
infrastructures into flexible connected EVs
to establish interactions with RSUs, LAGs, and
other entities [7, 8].
EV Edge Computing: The EVs act as distrib-
uted units to perform a substantial
amount of data processing and analysis
during collaborative operations, in which
sensors are geographical- ly dispersed to
establish interactions [9]. Flexible
computing is achieved instead of
establishing direct communications with
the remote cloud. This edge computing
mode provides higher accu- racy on
measurement, and emphasizes the prox-
imity to sensors, dense geographical
distribution, latency reduction, and
support for mobility to enhance the
quality of services [10, 11]. The mov- ing
EVs could enhance the network
connectivity and capability with high-
speed data transmission and low
communication latency.
This hybrid network architecture is
fundamen- tally different from that of the
traditional system paradigm. It is
established based on distributed
vehicular cloudlets and units, and
highlights the performance of local
information and energy pro- cessing with
the collaboration of nearby vehicular
resources.
VEHICULAR INFORMATION AND ENERGY INTERACTIONs
In the network structure, both information
and energy interactions (e.g.,
exchanging, collabora- tion, and
reallocation) are achieved to support the
Internet of Vehicular Energy.
Vehicular Information Interactions: During
information interactions, an EV
establishes wire- less communications
with multiple sensors for distributed
perception and cooperative process- ing,
including traffic updates monitoring and

IEEE Network • May/June 79

2018
 continually varies according to EVs’
arrivals and departures within an area
Internet of Vehicular [12, 13]. Meanwhile, an EV interacts with
Energy multiple sensors, and also supplies
redundant
RS energy to surrounding and
dispersed
U sensors based on the emerging
LA
wire- less charging technology for
G
creating a perpetual energy source to
provide power over distance, one-to-
many charging, and controllable wireless
power. Considering an EV’s activity cycle
per- spective, it moves around different
area networks along with distributed
energy support.
EV
E Note that
cloud4G LTE cellular telecommunication,
V WiFi, andcomputin
IEEE 802.11p/wireless access for
vehic- ular environments (WAVE) are
available broad- band wireless
EV communication technologies for vehicular
edge data transmission [14]. Both in-band
computin spectrum
Informationand the 5.9 GHz dedicated
g short-range
flow Energy communications (DSRC)
spectrum
flow are popular for ubiquitous
vehicular applications. The Inter- national
FIGURE 1. The network architecture of EVCE Standards Organization/International
computing. Electrotechnical Commission (ISO/IEC)
15118 standard is particularly designed to
support V2G communication interfaces.
municates with
fixed surrounding C ONTEXT -A WARE V EHICULAR APPLICATIONs
infrastructures, Figure 2 illustrates context-aware
communicates vehicular appli- cations, in which there are
with parking lots four scenarios referring to information and
to provide assis- energy interactions. The first two
tant information scenarios (shown by the red areas) refer
servers, and to
exchanges
messages with a
LAG during V2G
communications.
Vehicular Energy Interactions: During energy
interactions, an
EV performs
changing and dis-
charging
operations during
V2G
communications
with the roles of
energy demand,
energy stor- age,
and energy
supply. A LAG acts
as an agent
between the
power grid and
the gathered EVs
for aggregating
distributed
batteries.
The EVs own
potentially
available batteries
to interact with a
LAG for energy
exchanging. The
aggregated EVs
jointly establish a
large and flexi- ble
energy pool, and
the size of the
energy pool

80 IEEE Network • May/June

2018
 Virtual Power
Data Plants: During the
Small portable power calculators: cloud-com-
plants:energy + information + Virtual power
LA plants: energy puting-based
edge
G + cloud energy interactions,
the EVs act as
Senso virtual power plants
rs to provide flexible
RS energy aggregation
Network U
operators: with a cluster of
information +
scattered energy.
Discharging EV
Charging EV
Assume that there
Moving EV are many fully
Cloud charged EVs during
computing a long period of
Edge working time and
FIGURE 2. The contexts-aware vehicular applications based on the night- time, and the
information and energy interactions. EVs are potentially
available to feed
energy back into
information interactions, which consider the power grid or
the mov- ing EVs as infrastructures to other neighboring
enhance computation capability for the entities. Moreover,
vehicular cloud and edge. The other two an EV may play
scenarios (shown by the green areas)
refer to energy interactions, and involve
employ- ing the discharging EVs as
infrastructures for dis- tributed energy
support for the vehicular cloud and edge.
Note that the energy interactions are
performed along with the information
interac- tions, and the EVs are assigned
the following four main roles.
Spontaneous Network Operators: During the
cloud-computing-based information
interactions, slowly moving EVs act as
spontaneous network operators to
perform a substantial amount of
communications with neighboring EVs and
RSUs for establishing robust connectivity
[11]. The vehicular networking forms a
core component by carrying and
forwarding data packets to other EVs, and
is achieved through the combination of
V2I and V2V communications. The EVs
estab- lish interactions with the RSUs to
exchange infor- mation by monitoring
roadside infrastructures, and also
interacts with other EVs within a certain
range. Data-driven interactions provide
network connectivity for enabling data
transmission. Col- laborative networking
is achieved considering high mobility and
localized scattering from the neighboring
EVs and RSUs.
Mobile Data Calculators: During the edge-com-
puting-based information interactions, the
EVs act as mobile data calculators to
provide local data processing by
integrating various sensing data to extract
available information for transportation
man- agement. The moving EVs are traffic
probes, inte- grated with private sector
probe data for directly linking roadside
sensors to physical surroundings. For
instance, road detection and traffic
optimization of automated vehicles are
performed based on the cooperation of
proper sensors. The EVs approach network
boundaries with dense geographical distri-
bution and collaborative computing
capability.

IEEE Network • May/June 81

2018
 the role of mobile energy transporter to balance energy pools
among different areas.
Small Portable Power Plants: During the
edge-computing-based energy interactions, the EVs act as small
portable power plants to share energy with roadside
infrastructures (e.g., sensors and RSUs). Energy consumption of
sensors and RSUs is one of the most important constraints; it
will influence the network reliability and lifetime. Energy
connectivity is crucial for enabling energy flow among different
entities. The EVs are expect- ed to be applied for wireless
discharging to trans- mit idle energy to nearby sensors along
with the development of wireless charging technology. Long-
staying EVs become abundant and conve- nient power plants
for energy sharing.
sECURITY REqUIREMENTs AND DIsTRIBUTED
CONsENsUs sECURITY REqUIREMENTs
Traditional security requirements including the data
confidentiality, integrity, and availability (CIA triad),
authentication, authorization, and accounting (AAA), privacy
preservation (e.g., location, charge status, and identity) should
also be addressed. Due to the limitations of wireless
communication channels, data tampering, identity impersonation,
and privacy violation related secu- rity issues become severe in
EVCE computing.
Considering the characteristics of centerless trust,
collaborative intelligence, and spatio-tempo- ral sensitivity,
traceability and transparency should be highlighted.
Traceability: refers to data provenance to iden-
tify data lineage tracing for collaborative EVs and other entities.
It is required that the origins and intermediate flow of
information and energy be traced during interactions. Due to
attackers and legal entities being equally privileged and equi-
potent participants, the interactive data may be maliciously
utilized or tampering.
Transparency: refers to an entity (e.g., an EV)
knowing which other entity (e.g., a LAG) obtains its related data,
when and where the entity has used the data, and how the
entity realizes a spe- cific function. Transparency has limited
visibility during interactions based on the software defined
security model (e.g., zero-trust model), and pri- vate data
should be confused by scrambling mechanisms to avoid the
data being resolved by irrelevant entities.
In EVCE, there are dissimilar security challeng-
es.
Toward EV Cloud Computing: EVs establish information interactions
along with computation resource sharing, and it is challenging
to achieve data access control and traceability during dynam- ic
participation. EVs establish energy interactions with energy
resource sharing, and it is challenging to achieve aggregated
energy transmission with identity privacy preservation.
Toward EV Edge Computing: EVs communicate
with neighboring sensors, and it is challenging to achieve
anonymous data transmission and batch authentication. EVs and
LAGs establish direct ener- gy interactions, EVs and sensors
establish indirect energy interactions via RSUs, and it is
challenging to achieve centralized and distributed energy allo-
cation with energy status privacy preservation.

82 IEEE Network • May/June

2018
 Block Prev Hash Nonc Merkle Root Block Prev Hash Nonc Merkle Root
1 e 2 e

RS LA LA
U G G

Information Information Energy Energy

records records records records

Informatio Energy
n Data Energy
EVm coins coins
EVm EV EV
1 2 d c
Information Energy
interactions interactions
FIGURE 3. The structure of distributed consortium blockchain.

For instance, an EV performs and add them in a linear chronological order in a block- chain for
computations based on dispersive data, verification.
which will be further processed for a Figure 3 shows the structure of a consortium blockchain, and
certain purpose. Note that spatio- each block contains a cryp- tographic hash value to the prior
temporal attributes should be considered block. The traditional PoW in Bitcoin is available for the dis-
as criti- cal parameters, for which tributed consensus, which is achieved based on data contribution
timestamps and position information (e.g., frequency and energy contribu- tion amount.
latitude and longitude) could be jointly
applied for sequence recordings. Each
item of data should be assigned a lineage
flag for further identifying its origins.
Meanwhile, the EVs have dynamic
positions, which will be applied for
determining the location relationships for
the interactive entities, and the RSUs and
LAGs are regarded as having static
positions to assist loca- tion identification.
DIsTRIBUTED CONsENsUs
EVCE computing has similar features
(e.g., cen- terless trust and collaborative
intelligence) as the blockchain, in which all
the participants collectively validate new
blocks for collaborative management [15].
Blockchain establishes distributed
consensus before transaction records are
written into a digital ledger. It is executed
by the collaborative partici- pants based
on timestamps and Merkle hash tree
algorithms. The proof of work (PoW) and
proof of stake (PoS) are two typical
consensus algorithms. The PoW is
absolutely dependent on the comput- ing
power, and the participants compete to
obtain correct data writing with relatively
random and low probability. The PoS is
based on a deterministic approach and
probability, and an account is cho- sen
depending on its total stakes.
Here, data coins and energy coins are
defined
as new cryptocurrency for vehicular
applications. During information and
energy interactions, vehicular records are
stored in a consortium blockchain, and
the blockchain-inspired distrib- uted
consensus mechanisms are achieved. The
vehicular records will be encrypted and
structured into the blocks based on the
pre-defined distribut- ed consensus
mechanisms, and RSUs and LAGs
respectively audit the vehicular records

IEEE Network • May/June 83

2018
 Proof of Data sensors will not obtain data coins due to
Contribution their inherent functions for data
Frequency: During distribution. During the energy
information interactions, the sensors and RSUs will
interactions, data not obtain any energy coins since they
coins are defined have no redundant energy to perform
based on the discharging operations.
proof of EVs’ data The proof of data contribution
contribu- tion frequency and the proof of energy
frequency. A contribution are defined for determining
consensus process representation in majority decision
is performed by making. The established data coins and
authorized RSUs energy coins can be applied for vehicular
or LAGs for audit, resources allo- cation. If an EV contributes
and a new block more frequently for collaborative
will be formed for intelligence, it will obtain more data coins.
verification during It will be assigned higher priority to
a con- sensus access the resource pool, and its data
process. If the EV may be assigned higher credibility for
has the highest decision assistance. If an EV feeds more
data con- tribution energy back into the grid or other
frequency in a entities, it will own more energy coins,
certain period, it and will be assigned with higher priority
will be rewarded or lower price for energy utilization.
by data coins as
incentive to sECURITY sOLUTIONs FOR EVCE COMPUTING
encourage other In EVCE computing, there are moving EVs
EVs to contribute (i.e., EVm), discharging EVs (i.e., EVd),
information. charging EVs (i.e., EVc), RSU, LAG, and
Proof of Energy Contribution Amount: During multiple sensors. The data coins and
energy energy coins are considered during the
interactions, EVs’ interactions; anonymous data
energy coins are transmission
defined based on
the proof of EVs’
energy
contribution
amount, which is
directly related to
the periodic
energy
interactions. A
consensus
process is per-
formed by the
authorized LAGs,
and the amount of
discharged energy
is measured by
smart meters. If
an EV has the
most energy
contribu- tions in
a certain period, it
will be rewarded
by energy coins
for encouraging
other available
EVs to participate
in the discharging
operations.
The data coins
and energy coins
are only
exchanged among
EVs, and are
allowed to be cir-
culated and
traded. During the
information inter-
actions, the

84 IEEE Network • May/June

2018
 sharing and data
hiding issues
among dif- ferent
EV RS {EVm EVs.
U Energy-Driven
m i} Security Scheme: The
multi- ple
Key agreement and
distribution
discharging EVs
Access Access challenge {EVd1, EVd2, …,
challenge EVdj} (j ∈ N*) act as
Response and mutual authentication virtual power plants
Encrypted Anonymous data to establish an
data coin transmission Anonymous data aggre-
confirmation and gated energy
access control
resource pool via
the LAG. These
{EVdj} discharging EVs’
energy should be
LAG aggregated during
transmission with
Challenge and responses privacy
Multiple EVs' aggregated identities
considerations.
Authentication operators and

FIGURE 4. The security scheme in EV cloud computing.

and aggregated energy transmission are

achieved during heterogeneous
interactions.

sECURING INTERACTIONs IN CLOUD COMPUTING

Figure 4 shows secure interactions in cloud
com- puting, involving EVs (i.e., EVm and
EVd), an RSU, and a LAG, and the distributed
EVs’ resources are aggregated for more
flexible allocation and usage.
Information-Driven Security Scheme: The
moving EVs act as network operators to
estab-
lish V2V communications. EVm interacts
with its neighboring EVs {EVm1, EVm2, …,
EVmi} (i ∈ N*) for collaborative operations.
These moving EVs jointly perform
cooperations in which data-coin-
based anonymous data confirmation and
access control should be particularly
considered for data exchanging and
sharing.
In the initialization, the moving EVs
perform key agreement and distribution
based on the peer-to-peer networks;
temporary session keys could be
established based on lightweight sym-
metric encryption. The shortest path tree
routing and multi-path key mode could be
applied for group key agreement.
Thereafter, the moving EVs and the RSU
establish interactions via access chal- lenges
and responses. Here, the moving EVs joint-
ly complete cooperation for data
exchanging, the signed data could be
broadcast to neighboring EVs, and mutual
authentication could be estab- lished
based on homomorphic encryption and
secure multi-party computation.
The encrypted data coins directly
influence resource allocation among
different moving EVs. Moreover, spatio-
temporal attributes could be applied for
access control, and conditional proxy re-
encryption could be used to address data

IEEE Network • May/June 85

2018
 {EVd1, EVd2, …, EVdj} generate pseudo random numbers as
access challenges to be transmitted to the LAG for launching a
session. When the EVs and LAG establish interactions, the EVs’
aggregat- ed identifiers are transmitted to the LAG for mutu- al
authentication. The LAG extracts its local values to compute
authentication operators based on lightweight cryptographic
primitives for identifica- tion and authentication. Here, elliptic
curve digi- tal signature algorithm (ECDSA)-based signature
(e.g., ring, group, and blind signatures) could be applied for
anonymous data transmission.
The EVs have diverse energy preferences for
the discharging request. An EV supplies its idle energy back into
the power grid for aggregating distributed energy. During energy
interactions, the energy is regarded as a non-differential
resource for reallocation. The discharging EVs jointly estab- lish a
flexible energy pool, which can be shared by the EVs around the
same LAG. The Merkle- hash-tree-based selective disclosure
mechanism could be applied for protecting the sensitive data
fields. Such data structure avoids storing all the data fields to
realize efficient and secure verifica- tion of a large data
structure.
sECURING INTERACTIONs IN EDGE COMPUTING
Figure 5 shows secure interactions in edge com- puting,
involving the EVs (i.e., EVm, EVd, EVc), an RSU, a LAG, and sensors,
and the distributed EVs’ resources are applied for local computing
and processing.
Information-Driven Security Scheme: The moving EVm acts as a
mobile data calculator
during information interactions in edge comput- ing. Batch
authentication should be established by ultra-lightweight
algorithms.
The sensors constantly broadcast omni-bearing queries, and EVm
gives a response upon receiv- ing a periodic query. Thereafter, a
key agreement and distribution scheme could be established to
support dynamic participation and authentication. EVm and
sensors perform mutual authentication based on the
permutation, symmetric encryption, or hash/hash-based message
authentication code (HMAC). The multicast message
authentication and batch authentication become efficient for
secure interactions among multiple sensors.
Here, EVmfirst obtains the raw sensing data Data0, and further
performs processing and computing to obtain the advanced
data Data’0. Thereafter, EVm transmits Data’0 to an RSU for data
integration. Other neighboring EVs also per- form similar
authentication operations, and fur- ther transmit {Data’1, Data’2,
…, Data’x} to the RSU.
Based on the distributed consensus, Data’* will be written into
the digital ledger, and data coin will be assigned to the
appropriate EV.
Energy Driven Security Scheme: The discharg- ing EVd acts as a small
portable plant to establish
communications with EVc, LAG, RSU, and sensors during energy
interactions in the edge computing. For one aspect, the LAG
generates pseu- do-random numbers as access challenges,
and respectively transmits them to surround- ing EVc and EVd.
After the LAG and EVs estab- lish mutual authentication, EVd is
requested to perform discharging operation to feed a local
vehicle EVc. Upon EVd receiving the discharg- ing request, it
could wrap its energy status into

86 IEEE Network • May/June

2018
ciphertexts and transmit the encrypted
energy status to EVc. The delivery of
energy coins is performed based on Senso
EV
pseudonym, avoiding sen- sitive privacy m rs
disclosure. Sensors' periodical queries
For the other aspect, the sensors EV's responses
perform similar operations, and establish
Key agreement and
challenges and responses with EVd. The
sensors transmit ener- gy requests to EVd distribution Hash or HMAC-based
via an RSU for local energy access. EVd’s batch authentication
energy status will be transmitted to the
sensors for energy confirmation. Anonymous data transmission
EVc and sensors will be charged by the LA
wired and wireless charging technology, EV G EV
and energy coins are exchanged between c Challeng Challeng d
EVd and EVc based on the distributed e and e and
consensus. The concealed data response response
aggregation can be applied for privacy Mutual authentication Challenge and
preserva- tion, and hierarchical data Dischargin response Energy
access control can be applied intra- g
reques request
network (e.g., V2G communications) and t
inter-networks (e.g., V2G and V2I
communi- cations). Encrypte
d energy Energy Energy
In EVCE computing, the data coins and coin confirmatio confirmation
ener- gy coins are applied as n
authentication operators for EVs. The
access priority, credibility, and price could FIGURE 5. The security scheme in the EV edge computing.
also be related to data coins and energy
coins. Data traceability and transparency
.
could be achieved by the inherent
characteristics of block- chain.
CONCLUsION
This article focuses on security issues for
both information and energy interactions in
EVCE com- puting. The context-aware
vehicular applications are presented
according to the EVs’ different roles;
blockchain-inspired data coins and energy
coins are defined to achieve distributed
consen- sus; and data contribution
frequency and ener- gy contribution
amounts are applied for proof
determination. Security schemes are
proposed for cloud and edge computing to
launch perspectives on vehicular
applications.
ACKNOWLEDGMENT
This work is funded by the National
Key R&D Program of China
(2017YFB0802302, 2017YFB0801701)
and the National Natural Sci- ence
Foundation of China (61601129). This
work is partially supported by projects
240079/F20 funded by the Research
Council of Norway.

IEEE Network • May/June 87

2018