BlockJack: Towards Improved Prevention of IP Prefix Hijacking Attacks
in Inter-domain Routing via Blockchain
I. Wayan Budi Sentana
a
, Muhammad Ikram
b
and Mohamed Ali Kaafar
c
Department of Computing, Macquarie University, 4 Research Park Drive, Macquarie Park, Sydney, Australia
Keywords:
Blockchain, Border Gateway Protocol, Prefix Authorization, Prefix Hijacking, Prefix Verification.
Abstract:
We propose “BlockJack”, a system based on a distributed and tamper-proof consortium Blockchain that aims
at blocking IP prefix hijacking in the Border Gateway Protocol (BGP). BlockJack provides a synchroniza-
tion amongst a BlockChain and BGP networks through interfaces ensuring operational independence. This
approach preserves the legacy system and accommodates the impact of a race condition if the Blockchain
process exceeds the BGP update intervals. BlockJack is also resilient to dynamic routing path changes during
the occurrence of the IP prefix hijacking in the routing tables. We implement BlockJack using Hyperledger
Fabric Blockchain and Quagga software package and we perform an initial set of experiments to evaluate
its efficacy. We evaluate the performance and resilience of BlockJack in various attack scenarios including
single and multiple paths attacks, and attacks from random sources. Our results show that BlockJack is able to
handle multiple attacks caused by Autonomous Systems (AS) path changes during a BGP prefix hijacking. In
experiment settings with 50 random routers, BlockJack takes on average 0.08 seconds (with standard deviation
of 0.04 seconds) to block BGP prefix hijacking attacks.
1 INTRODUCTION
Border Gateway Protocol (BGP)–also known as the
Inter-domain routing protocol–is a path-vector proto-
col that regulates the connectivity and information ex-
change among Autonomous Systems (AS).
1
Based on
data presented in APNIC Research and Development
(AS 65000), there are currently almost 70,000 unique
ASN seen in their BGP routing table (BGP-Potaroo,
2020). Each AS maintains a number of IP Prefixes (in
short Prefixes) and domains assigned by the Internet
Assigned Number Authority through Regional Inter-
net Registries (RIR).
To hijack an IP prefix, an adversary router (or
AS) advertises a (fake) IP prefix that belongs to an-
other router (or AS). When the adversary AS con-
ducts a prefix advertisement, BGP sends the prefix
to all neighbours on the Internet. As a result, the
traffic that is supposed to reach the original AS, is
redirected to the adversary AS which may result in
a
https://orcid.org/0000-0003-3559-5123
b
https://orcid.org/0000-0003-0336-0602
c
https://orcid.org/0000-0003-2714-0276
1
An AS is an independent network that comprises the Inter-
net and each AS assigned a 16-bit or 32-bit unique number
known as Autonomous System Number (ASN).
unavailability of Internet services for the targeted AS
and other breaches of the security of the Internet traf-
fic. Besides adversarial considerations, instances of
IP Prefix hijacking can be due unintentional network
mis-configurations (e.g. during network setup (Hope,
2020)).
To prevent BGP hijacking and to enable the val-
idation of the origin of BGP announcements, IETF
(the Internet Engineering Task Force) supports the
Resource Public Key Infrastructure (RPKI) through
RFC 6480. IETF further released RFC 6482 for Route
Origin Authorization (ROA) and RFC 6483 for Route
Origin Verification (ROV). ROA is a process where
an AS authorizes a number of prefixes to be adver-
tised under its jurisdiction, and stores it as a tuple of
IP prefix, AS (owner), the maximum length of AS and
expiry date of each IP block (Iamartino et al., 2015).
To prevent prefix hijacking, the tuple can be utilized
during ROV process to verify whether or not an AS
advertises the authorized prefix.
The RPKI architecture gradually delegates the
process of securing BGP from Internet Assigned
Numbers Authority (IANA) as the global regulator
to Regional Internet Registrar (RIR), Internet Service
Providers (ISPs) and private network companies (Liu
et al., 2016). Each of these institutions is allowed to
674
Sentana, I., Ikram, M. and Kaafar, M.
BlockJack: Towards Improved Prevention of IP Prefix Hijacking Attacks in Inter-domain Routing via Blockchain.
DOI: 10.5220/0010521106740679
In Proceedings of the 18th International Conference on Security and Cryptography (SECRYPT 2021), pages 674-679
ISBN: 978-989-758-524-1
Copyright
c
2021 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
publish a Certificate of Authority (CA) to its down-
stream network and keep the Resource of Certifi-
cate (RC) in its storage. This process is considered
to be rather risky as any attack or a potential miss-
configuration on the upstream network can cause the
failure of the entire downstream network. This hier-
archical structure also provides the privilege for the
upstream network to overwrite the RC of its down-
stream network (Cooper et al., 2013).
We propose BlockJack, a consortium
blockchain-based system to prevent IP prefix hi-
jacking in Border Gateway Protocol. Instead of
providing a centralized repository as in RPKI,
BlockJack introduces a distributed repository of IP
prefix-ASN pairs in consortium blockchain nodes for
prefix authorization and prefix verification purposes.
The BlockJack system also stores the consortium
member’s CA in each AS repository to eliminate the
drawback of a centralized or hierarchical public key
infrastructure-based scheme.
BlockJack independently runs Blockchain along
with BGP network. It creates an interoperable module
to synchronise the operations of the Blockchain and
BGP network thus, unlike (Liu et al., 2020), Block-
Jack eliminates the need for router software update.
The architecture of BlockJack provides reliability by
avoiding the race conditions between the BGP Update
Message
2
and the Blockchain process, such as prefix
authorization or prefix verification. A race condition
occurs when the Blockchain transaction runs inside
the BGP protocol loop and exceeds the BGP Update
Message interval.
To improve robustness of BlockJack’s to the dy-
namic nature of BGP AS Path
3
changes and AS path
divergence–oscillating AS Path changes during the
prefix hijacking, BlockJack stores only the prefix and
AS origin in the Blockchain. BlockJack uses Next
Hop
4
information to identify the neighbouring AS
that contributes to the addition of prefix in the routing
table, in order to create the Inbound filter. In case of
a hijacking event, an inbound filter blocks the incom-
ing Prefixes from a certain AS through the immediate
neighbors. This approach reduces the number of ver-
ification and authorization process to the Blockchain
caused by dynamic change of AS Path.
The rest of this paper is organized as follows. Sec-
2
BGP Update Message is a signal used to refresh the con-
tent of BGP routing tables. The default Update message
interval for Cisco router is 30 seconds (Vinit and Brad,
2018).
3
AS Path is a parameter that stores the list of AS and use to
retrieve the prefix source.
4
Next Hop is a parameter used to identify the immediate
router in a given AS Path.
tion 2 describes the architecture and mechanisms that
support BlockJack operations. Section 3 describes
Test-bed, BlockJack’s test scenarios and analysis of
our experiments. While Section 4 presents related
work and Section 5 concludes our research.
2 SYSTEM ARCHITECTURE
AND MECHANISM
In this section, we present the architecture, process
and mechanism of BlockJack.
2.1 System Architecture
BlockJack consists of the following three main mod-
ules as depicted in Figure 1:
Blockchain. To regulate the interaction among
ASes demanding a highly trusted environ-
ment, we leverage Hyperledger Fabric (Linux-
Foundation, 2020) to build the Blockchain
module. Unlike public-based (permissionless)
Blockchain (such as Bitcoin or Ethereum), Hy-
perledger Fabric eliminates the role of miners in
tethering the new blocks to the existing blocks
by adding several components.
5
The consen-
sus mechanism ensures only trusted and known
consortium members to participate in Blockchain
transactions.
Profiler. This module consists of the process for
generating the private key, public key, and X.509
certificate for admin and routers in each AS. To
store these credentials, Profiler offers a wallet and
provides a REST based API to share the creden-
tials among different (sub)modules of BlockJack.
Dispatcher. This module conducts routine tasks
to monitor the routing tables and dispatches filter-
ing commands if there is any updates about the
BGP routing tables.
2.2 BlockJack’s Mechanisms
BlockJack consists of supporting (control) and main
mechanisms. Supporting mechanism operates the
Blockchain network, creates credentials, initializes
the Blockchain, and registers an admin router for con-
trol and monitoring purpose. In the following, we
5
Hyperledger Fabric provide Orderer, Commiter, Chain-
code, Endorser and Fabric CA to support the blockchain
transaction mechanism known as consensus (Linux-
Foundation, 2020).
BlockJack: Towards Improved Prevention of IP Prefix Hijacking Attacks in Inter-domain Routing via Blockchain
675
Figure 1: An overview of the three modules of BlockJack.
The Blockchain module is handling data storage and data
query while Profiler acts as a bridge between Blockchain
and Dispatcher as well as to store the Credentials of each
router. Dispatcher monitors routers and dispatches filter
commands if Prefix Hijacking occurs.
provide more details about the main mechanisms in
BlockJack:
Prefix Authorization. This mechanism validates
and proves that the prefix belongs to specific
ASes. To claim the ownership of a Prefix, an AS
first starts the Dispatcher module to read the BGP
routing table and captures the valid-best route
from the prefixes list. If a new prefix is found,
Dispatcher sends an HTTPS request to the Pro-
filer asking for prefix addition to the Blockchain.
In the Profiler, the REST API server receives the
request and authenticates the request to the router
profile. The Profiler then adds the router creden-
tials to the request and sends it to the Blockchain.
In the Blockchain, Endorser verifies the request
based on the contract written in the Chaincode.
Once the request complies with the smart con-
tract, the Endorser sends the prefix to the Orderer
and then initiates the consensus mechanism in
the Blockchain. The Orderer module distributes
the requests to all consortium members and gets
the approval from at least 50% of the consortium
members. When consensus is reached, the Or-
derer signals the Committer to seal the transaction
and adds a new block into the Blockchain ledger.
At the end of the Blockchain transaction, the Or-
derer creates a new World State
6
and distributes
it to the whole blockchain nodes.
Prefix Verification. Prefix Verification module
aims to check whether the announcement received
by the router from its neighbors contains the au-
thorized prefix. For this purpose, BlockJack
compares the prefixes found in the routing table
with the prefixes stored in the Blockchain. Pre-
6
World State is a current state of database (Ledger) in
blockchain.
fix Verification mechanism starts when the Dis-
patcher Module finds a new prefixes announce-
ment from the router’s neighbor, and then requests
for verification to the Blockchain through Pro-
filer. In the Profiler, Rest API server conducts
authentication against the router credentials, and
then sends the request to the Blockchain. In the
Blockchain, the Chaincode verifies the compli-
ance of the request and then query the Ledger.
The query result publishes three kinds of signals;
Valid, Invalid, and Unknown. The Valid signal in-
dicates that the prefix sent to the Blockchain is
available and corresponds to the legitimate AS
Number. While, an Invalid signal denotes that
the prefix exists in the Ledger but corresponds to
different AS Number. This invalid signal can in-
dicate that a BGP prefix hijacking occurs in the
network. The Unknown signal indicates neither
the prefix nor the AS Number are available in the
Ledger. When the Dispatcher accepts the Invalid
signal, then it produces Inbound filter commands
to block the announcement from the AS that was
indicated as a source of the prefix hijacking.
3 TESTBED SETUP AND
EVALUATION
This section presents our experimental testbed and
analysis.
3.1 Experiment Testbed
Blockchain module of BlockJack is implemented us-
ing Hyperledger Fabric. Each Blockchain node, in-
cluding Orderer, Fabric CA, and Chaincode are run-
ning in a separated Docker container. For the Profiler
module, we tailor Node JS to create the admin func-
tion, router registration, and REST API server. Using
Python and shell script, we develop the Dispatcher
module. While develop BGP network environment
leveraging Quagga router software, and for this re-
search, we use Dockerized Quagga image produced
by (Chiodi, 2020). In our testbed, each router runs in
a separate detach-mode Docker container connected
by a customized virtual network. We use the testbed
and conduct several sets of experiments to evaluate
the performance and resiliency of BlockJack against
BGP hijacking attacks. We provide further details in
the following:
Performance Evaluation. We aim to evaluate the
performance of our BlockJack in terms processing
time required for handling prefix authorization and
verification requests. To this end, we conduct two
SECRYPT 2021 - 18th International Conference on Security and Cryptography
676
sets of experiments
7
and record the processing time
of the authorization and verification mechanisms of
BlockJack. In particular, we generate sets of ran-
dom prefixes from a BlockJack router and query the
Blockchain ledger to determine the authorization and
verification times required by our proposed systems.
For the prefix authorization setup, we create a
function in the Dispatcher module to send a various
number of prefixes into the Blockchain to measure
BlockJack performance in handling prefix authoriza-
tion. Each prefix authorization process followed by
commit order so each prefix addition will be adding a
block in the Blockchain. At the end of the experiment
there will be a thousand blocks in the chain. To mea-
sure the prefix verification time, we create a script so
the Dispatcher can send a various number of prefixes
to be verified by Blockchain with 1000 blocks.
Resiliency Evaluation. Resiliency evaluation aims
at observing the BlockJack’s neutralizing capabili-
ties against the prefix hijacking attacks. To measure
the resiliency of BlockJack, we leverage Quagga and
Docker in the higher computing environment
8
and
create several network topologies with various num-
ber of routers. We measure the prefix hijacking neu-
tralization in two stages including Prefix Prepend-
ing and Neutralization. Prefix Prepending is the pro-
cess of adding an ASN to the AS-path parameter
in the BGP table for each AS passed by a prefix.
In this experiment, the prepending time is equal to
the time needed by the adversarial prefix to arrive at
the router where the Dispatcher resides (BlockJack
router) and disrupts the original prefix as the route
with the valid-best status. While, Neutralization is
the stage where BlockJack detects, verifies, and sends
filter commands to the router to neutralize hijacking.
By measuring the prepending (BGP hijacking attack)
time and neutralization (blocking) time, we determine
the duration of BGP hijacking attacks and the effi-
ciency of BlockJack to neutralize the attacks, respec-
tively.
For this experiment, we create the following three
different scenarios as:
Single Path Attack Scenario. We create a bi-
nary tree-like network topology and deploy the
dispatcher on the router which is located at the
root. We prepared five adversarial prefixes that
would be used by routers located in the farthest
branch to hijack the prefixes announced by routers
in the leaf of the tree. These attacks create single
paths when they reach the BlockJack router i.e.,
7
Experiments were performed on a workstation with
Ubuntu 18.04, CPU 2.7GHZ, and RAM 16GB.
8
Experiment conduct in Cluster Server with 4 core CPU,
128 GB of memory, 500 GB HD running on Ubuntu 18.04.
root of the tree.
Multiple Path Attack Scenario. This scenario is
designed to examine BlockJack’s resilience in an-
ticipating routing path changes that occur during
BGP prefix hijacking. In this scenario, we modify
the binary tree network topology in the first sce-
nario by setting up BGP peering for each branch
at the same level. This will cause each announced
prefix to have more than one path when it reach
BlockJack router at the root of the tree.
Random Attack Scenario. This scenario is set
up to examine the BlockJack resilience in a very
random BGP environment. We made several ran-
dom network topologies with various numbers of
routers. The connectivity level in each experiment
was set to 25% indicating that a node has a prob-
ability to be connected to 25% of the total nodes
on the network. For each set of experiments, for
realistic evaluation, we randomly place the Dis-
patcher and five random adversarial prefixes in
our testbed.
Limitations. We collect 25 experimental results for
each scenario in networks consisting of 20 to 60
routers. We observe that our testbed suffers with the
increase in number of routers. Docker network shares
the same Linux kernel to handle all virtual networks
running on top of the Docker container. This condi-
tion causes an overload on the virtual network. It also
causes the Hyperledger failure to install the Chain-
code (smart contract) on the Blockchain node through
the corresponding port.
Table 1: Time recorded by BlockJack to authorize and ver-
ify prefixes. Each prefix addition followed by commit pro-
cess and create a new block.
# of Pref. Authorization Verification
(Block) Avg. (s) Total (s) Avg. (s) Total (s)
100 2.16 216.21 0.1 10.27
500 2.15 1,076.61 0.09 47.38
1,000 2.15 2,154.32 0.09 92.12
3.2 Analysis
Performance Analysis. This Section presents, the
performance analysis of BlockJack in terms of time
required for prefix authorization and verification:
Prefix Authorization Analysis. Table 1 shows pre-
fix authorization time increase gradually according to
the number of prefixes sent to the Blockchain. On
average, BlockJack needs 2,154.32 seconds to autho-
rize 1,000 prefixes with an average authorization time
of 2.16 seconds. This heavy or expensive process is
caused by a complex consensus mechanism during
BlockJack: Towards Improved Prevention of IP Prefix Hijacking Attacks in Inter-domain Routing via Blockchain
677
(a) Single Path Attack. (b) Multiple Path Attack. (c) Random attack.
Figure 2: Resiliency evaluation results. The prepending time record increases gradually as the number of routers increases
in single path and multiple path attack scenarios. While, the prepending time for random attacks fluctuates. All prepending
times are presented in the form of time / 10 seconds.
the block addition to the chain.
Based on that result, if the BGP update interval
is 30 seconds, the maximum prefixes that can be au-
thorized in one interval is 13 prefixes which is higher
than 12 prefixes average announcement per AS origin
as shown in (Tony Bates, 2020). Hence, in this case,
BlockJack can handle the average prefixes announce-
ment in real world condition in one BGP update mes-
sage interval simultaneously without the assistant of
the local ROA cache, disregarding the network traffic
delay.
We then compare this prefix authorization result to
the real condition of the prefix ownership. According
to (Tony Bates, 2020), the highest number of prefixes
announced by an AS is recorded by AS8151 (Uninet
S.A. de C.V., MX) with 8125 prefixes. In this case,
BlockJack needs 625 BGP update interval or 17,550
seconds to authorize the whole prefixes when it runs
for the first time which potentially creates a race con-
dition between BGP Update interval and prefix au-
thorization if the Blockchain node embeds inside the
router machine and depends on the BGP message
signal. The time taken to access Blockchain during
the authorization process far exceeds the BGP update
message interval. Hence, the Blocjack’s decoupling
of Blockchain and routing environment helps in re-
solving any potential race-conditions.
Prefix Verification Analysis. The experiment re-
sult in Table 1 shows that the prefix verification pro-
cess is much lighter than that of prefix authorization.
For instance, BlockJack needs 92.12 seconds to verify
1,000 prefixes or on average of 0.09 second per prefix.
Given the worst case scenario that the growth projec-
tion of 150 prefixes per day (Geoff, 2020) appear in
the concurrent time, and assume that the same amount
of prefix withdrew in the same time, then BlockJack
only need 27 seconds to verify the 300 updated pre-
fixes. This record is below the 30 seconds BGP Up-
date message interval. Hence, in this case Block-
Jack can handle Prefix verification without local ROV
cache assistance.
However, verifying the global BGP routing table
would be so challenging to conduct in one BGP up-
date message interval. Expect that the number of the
global routing table is 850,000 (BGP-Potaroo, 2020)
and assume that the verification time per prefix is
0.09 seconds, then the total time needed by Block-
Jack to verify those prefixes on the first time running
is 76,500 seconds. That result is equal to 2,550 of
the BGP UPDATE message interval. Hence, the ex-
istence of a local ROV cache is crucial to reduce the
verification request during the BlockJack operation.
The Dispatcher can compare the entry of BGP rout-
ing table and the local ROV cache to find a new prefix
announcement or withdrawal occurrences, and then
verify those prefixes. As an addition, any verification
that exceeds the BGP message interval is not impact-
ing the BlockJack because the Dispatcher is not de-
pendent on any BGP message signal.
Resiliency Analysis. BlockJack is able to neutralize
all adversarial prefixes that disrupt the BGP routing
table in the three attacking scenarios. The results of
BlockJack’s resilience evaluation are depicted in Fig-
ure 2. We observe that the average prepending time
increases gradually in accordance with the router ad-
dition in single path and multiple path attack scenar-
ios, while the average prepending time for random at-
tacks seems to fluctuate. The lowest prepending times
for each single path, multiple path and random at-
tack scenario were 28.068, 41.855, and 52.101 sec-
onds, which are recorded during the experiment with
20 routers. The average prepending time for all exper-
imental sets on single path, multiple paths and random
attack scenarios, respectively, is 74.527, 80.9088, and
54.9572 seconds.
Neutralization time looks constant in all scenarios
with an average of 0.1516 seconds in single path sce-
narios and 0.2362 seconds in multiple path scenarios,
except for random attack scenarios which record an
average neutralization time of 1.0484 seconds. The
SECRYPT 2021 - 18th International Conference on Security and Cryptography
678
amount of neutralization time in random attack sce-
narios is up to 5 times compared to multiple path
scenarios. This is because the number of neighbors
in the random attack scenario is greater than multi-
ple path scenarios. From the five adversarial prefixes
sent, the average number of attacks on the random
scenario reaches 10.08 attacks, compared to 4.52 at-
tacks received by BlockJack routers in multiple path
attack scenarios. If we take a sample of random attack
scenario with 50 routers, on average BlockJack needs
0.957 seconds to neutralize 12.04 attacks of ve ex-
periment attempts. That record is equal to 0.08 sec-
onds to neutralize a single attack.
The standard deviation of prepending and neu-
tralization time is also seen constant in single path
and multiple path scenarios with an average range
of 8.2488 seconds to 11.6154 seconds for prepending
time and 0.0162 seconds to 0.0188 seconds for neu-
tralization time. While the average standard deviation
of prepending and neutralization time in the random
attack scenario was recorded at 20.3712 seconds and
0.8998 seconds, respectively.
4 RELATED WORK
The closest work to our research is the work proposed
by Sfirakis et al., (Sfirakis and Kotronis, 2019). In
the paper Sfirakis et al., introduced the concept of
a Blockchain-based prefix hijacking prevention us-
ing Bitcoin. The proposed system has been tested
in Quagga router software. However, this system re-
quired the AS owner to provide share coins for every
prefix authorization or prefix verification requests, as
the blockchain need miners to attach new blocks to
the blockchain.
5 CONCLUSION
Although the Prefix Authorization and Prefix Ver-
ification processes can ideally be handled in one
BGP Update message, several conditions will cause
a race condition between processes that occur on
the Blockchain and processes that occur in BGP. In
this paper we presented how Blockjack addresses this
specific issue. BlockJack also manages dynamic-
multiple hijacking scenarios due to changes in the
BGP attribute values which cause dynamic changes
in determining the best-valid path in the BGP routing
table.
ACKNOWLEDGEMENTS
The first Author is a Scholarship Awardee of the
Indonesian Endowment Fund for Education (LPDP)
of the Republic of Indonesia, with the LPDP ID
20193221014024. This research is partially funded
by the Optus Macquarie University Cyber Security
Hub.
REFERENCES
BGP-Potaroo (2020). Bgp analysis report-bgp table. https://
bgp.potaroo.net/index-bgp.html. BGP Table Data last
accessed 1 July 2020.
Chiodi, P. C. (2020). Quagga router software code
@github. https://github.com/pierky/dockerfiles/tree/
master/quagga. last accessed 1 July 2020.
Cooper, D., Heilman, E., Brogle, K., Reyzin, L., and Gold-
berg, S. (2013). On the risk of misbehaving rpki au-
thorities. In Proceedings of the Twelfth ACM Work-
shop on Hot Topics in Networks, HotNets-XII, New
York, NY, USA. Association for Computing Machin-
ery.
Geoff, H. (2020). Bgp in 2019 – the bgp table. https://blog.
apnic.net/2020/01/14/bgp-in-2019-the-bgp-table/.
Blog APNIC last accessed 1 Juli 2020.
Hope, A. (2020). Russian rostelecom compro-
mises internet traffic through bgp hijack-
ing. https://www.cpomagazine.com/cyber-
security/russian-rostelecom-compromises-internet-
traffic-through-bgp-hijacking. Accessed: 18/12/2020.
Iamartino, D., Pelsser, C., and Bush, R. (2015). Measur-
ing bgp route origin registration and validation. In
Mirkovic, J. and Liu, Y., editors, Passive and Active
Measurement, pages 28–40, Cham. Springer Interna-
tional Publishing.
Linux-Foundation (2020). Hyperledger fabric release
2.0. https://hyperledger-fabric.readthedocs.io/en/
release-2.0/whatis.html. Hyperledger Fabric Release
2.0 readthedocs last accessed 1 July 2020.
Liu, X., Yan, Z., Geng, G., Lee, X., Tseng, S.-S., and Ku,
C.-H. (2016). Rpki deployment: Risks and alternative
solutions. In Zin, T. T., Lin, J. C.-W., Pan, J.-S., Tin,
P., and Yokota, M., editors, Genetic and Evolutionary
Computing, pages 299–310, Cham. Springer Interna-
tional Publishing.
Liu, Y., Zhang, S., Zhu, H., Wan, P.-J., Gao, L., Zhang, Y.,
and Tian, Z. (2020). A novel routing verification ap-
proach based on blockchain for inter-domain routing
in smart metropolitan area networks. Journal of Par-
allel and Distributed Computing, 142:77 – 89.
Sfirakis, I. and Kotronis, V. (2019). Validating IP
prefixes and as-paths with blockchains. CoRR,
abs/1906.03172.
Tony Bates, Philip Smith, G. H. (2020). Cidr report for 1
jul 20. https://www.cidr-report.org/as2.0/. CIDR Data
report last accessed 1 July 2020.
Vinit, J. and Brad, E. (2018). BGP Message. Cisco Press,
San Fransisco.
BlockJack: Towards Improved Prevention of IP Prefix Hijacking Attacks in Inter-domain Routing via Blockchain
679