IEEE Consumer Electronics Magazine - September/October 2021 - 90

AI-Driven Security Solutions for the Internet of Everything
if the training data as well as testing data fall into
different domains, the correctness of the object
classifier is also reduced significantly. In recent
years, various algorithms have been introduced
to mitigate the domain shift data sets. However,
the existing evaluation techniques for the
domain adaptation depict a closed set domain
adaptation, where both the source and target
domains contain identical object classes. As
mentioned before, the poisonous data sets contain
various domain data, and therefore, the
closed set domain adaptation will not be helpful
for well prediction in this scenario. As a result,
an open set domain adaptation has been initiated,
which has more pragmatic scene where
only a few categories of data are shared among
the source and target data. In such a case, both
the target and source domains hold several
data, which are not part of interested classes. In
other words, the target domain may contain various
data, which are not relevant to any data in
the source domain and vice versa. Moreover, by
adopting the open set domain adaptation, it
could be easy to identify the poisonous data by
separating the data into a different class,
because this type of data does not belong to any
predefined classes.
Now, the transactional data stored in the
blockchain are genuine and authentic as they
are verified and added in the blockchain
through the voting-based consensus mechanism.Assumethatwewant
toperform thebig
data analytics on the transactional data in the
blocks. The big data analytics on such genuine
and authentic data are helpful to the AI/ML
algorithms to work as per their expectations
as the data set is not malicious due to the
absence of the data-poisoning attacks. As a
result, the correct predictions on the blockchain
data will help the businesses and organizations
to keep the financial terms and their
reputations as well. This process has been
illustrated in Step 12 (see Figure 4). Sometimes,
a user needs to access the data that are
stored in the blockchain. For this purpose, the
user requires to send a data access request to
the cloud server, say CSn. Once the cloud
server verifies the user, it grants access to the
data containing in the blocks, which are public
in nature. The user then verifies the requested
90
blocks, and accesses the data from those
blocks once the blocks verification is successful,
as showninStep11(seeFigure4).
BLOCKCHAIN IMPLEMENTATION
In this section, implementation of the blockchain
has been done with the system configuration
having the setting: Ubuntu 18.04, 64-b OS
with Intel(R) Core(TM) i5-4210U CPU @ 1.70 GHz,
4 GB RAM, and it has been developed in the
Node.js language with VS CODE 2019.
To computethesizeof apublicblockprovided
in Figure 5, we take block version, previous block
hash, Merkle tree root, block class, timestamp
(epoch time), block owner, owner public key (ECCbased
public key), block payload, current block
hash ðCBHashÞ (using SHA-256 hashing algorithm),
and block signature (ECDSA signature on CBHash)
of sizes 32, 256, 256, 32, 42, 160, 320, 160nt (where nt
is the number of transactions stored in a block),
256,and 320b.Therefore,the totalsizeofapublic
block becomes 1674 þ 160nt b.
Similarly, one can also calculate the size of a
private block in Figure 5 with the same assumptions;
apart from only difference is that the
encrypted block payload needs 640nt b, as each
transaction Txi in the block payload is encrypted
using ECC encryption, which outputs two elliptic
curve points requiring ð320 þ 320Þ¼ 640 b. Thus,
the total size for a private block becomes 1674 þ
640nt b. To compute the size of a consortium
block in Figure 5, we consider nt=2 encrypted
transactions and nt=2 unencrypted transactions
in the block payload. Therefore, the total size for
a consortium block becomes: 1674 þ 640ðnt=2Þþ
160ðnt=2Þ¼ 1674 þ 400nt b.
The simulation has been performed over the
public blockchain scenario. Similarly, one can
do the blockchain simulation for other cases of
private and hybrid blockchains. The number of
peer nodes in the P2P cloud server ðCSÞ network
is taken as 10. A fog server ðFSÞ securely sends
the data to the associated cloud server in the
form of transactions. CS then gathers the transactions
and put them into a transaction pool. If
the number of transactions in the transactions
pool reaches to a predefined transaction threshold
(i.e., the minimum number of transactions
needs to store in a block), a leader is selected
from the network for generating a block and
IEEE Consumer Electronics Magazine
IEEE Consumer Electronics Magazine - September/October 2021

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - September/October 2021
