2019 Fifth International Conference on Image Information Processing (ICIIP)

Implementation of Distributed File Storage and

Access Framework using IPFS and Blockchain
Randhir Kumar Rakesh Tripathi
Department of Information Technology Department of Information Technology
National Institute of Technology,Raipur National Institute of Technology,Raipur
CG, India-492010 CG, India-492010
rkumar.phd2018.it@nitrr.ac.in rtripathi.it@nitrr.ac.in

Abstract—Many critical applications are designed on the BitTorrent is that , about 170 million users uses BitTorrent
distributed structure using the blockchain technology to every month [5]. Moreover, BitTorrent protocols takes 40%
ensure the availability, immutability, and security. However, of the internet traffic on daily basis [5]. However, these ap-
these applications are facing the storage problem owing
to the data volume growth of transaction. The number plication does not secure and efficient when compared with
of transactions and its size in a block is growing in the HTTP. The HTTP protocol is one of the most considerable
blockchain day by day because of the feature of immutability protocol in terms of development for distributed applications
and append-only. The growing nature of transactions in worldwide. However, the HTTP protocols work on location-
a block is not only making the problem for storage but addressed based protocols which is not much efficient in
also in access to the block transactions. In this paper, we
propose an IPFS based blockchain storage model to solve the terms distributed access of file. To overcome the drawback of
storage problem of transaction in a block along with access BitTorrent and HTTP , the Interplanetary File System(IPFS)
of transaction of a particular block. In the propose storage [3], distributed file storage network is used for building a
model, the miners stores transaction on IPFS distributed file better web for us. The IPFS works with the content-addressed
system storage and get the returned IPFS hash of transaction technique to store and access of file.
into the block of the blockchain. The feature of the IPFS
network and its resultant hash reduce the size of transactions The distributed storage provides structure where data
in a block. To secure access of transaction for a particular is disseminated on different nodes, which refers the same
block content-addressed (IPFS hash) storage technique has distributed storage system. The each node is peers with
been proposed. We have applied this scheme on a transaction one another in the distributed network. The all peers keeps
which includes image storage on IPFS and hash storage same capabilities to share and access the files i.e., no one is
into the blockchain. In this paper, we have also proposed
the content-addressed technique in contrast to the location important and superior to others. Moreover, the distributed
addressed for the access of transaction. To implement the network also replicates the information which ensures the
framework we have used anaconda python, python flask, availability of information in the system.
and IPFS. IPFS(Interplanetary File System) provides distributed file
Index Terms—Blockchain, InterPlanetary File System(IPFS), storage system which facilitates connection with P2P net-
Distributed storage, Content-addressed technique.
work. The IPFS calculates the unique Hash of a file which
is accessible to all the peers in the network. The hash get
I. Introduction
modified every time when file gets updated [6].
Recently, the blockchain technology is emerging as a The IPFS is considered as the backbone of web 3.0 which
decentralized database storage technology [1]. Moreover, provides P2P decentralized file storage system and content-
the blockchain provides features such as immutable, shared, addressable technique to access the stored file [3]. IPFS
and decentralized database that stores history of assets and facilitate the distributed hash table (DHT) which is much
transactions across a distributed peer-to-peer network. The efficient than BitTorrent, and Git file storage system. IPFS
blockchain is successfully utilized in bitcoin developed by is also known as version controlled system which ensure
Satoshi Nakamoto but now it is gaining popularity in the the security, reliability, and scalability faced by the existing
various fields including healthcare, supplychain, and finan- file storage and sharing system. The same transaction has
cial sector [2]. the same hash in the IPFS that ensures the originality of
There are many distributed file storage applications cur- the content. The IPFS hash gets disseminated to all the
rently running all over the world. However, these applica- peers which also ensures the consistency among the peers.
tions are having problem of storage space and access with Interplanetary File system provides high throughput content-
other peers in the network. The BitTorrent is the distributed addressed block storage model which ensures security of
file storage and sharing application which provides the transactions.
facilities to coordinate with untrusted peers of the distributed IPFS provides a high throughput in terms of access using
network [4]. The more important and notable thing of the content-addressed block storage model, we also notice that

978-1-7281-0899-5/19/$31.00 ©2019 IEEE 246

 2019 Fifth International Conference on Image Information Processing (ICIIP)

storage is a new challenge in blockchain technology owing to transaction data in blockchain. The transaction data (image
the append only features [4]. In this paper, we try to leverage file) is stored into the IPFS. Only the IPFS hash data is stored
the concept of decentralized P2P system in IPFS to eliminate as a transaction in the block of blockchain [14].
the limitations of blockhain storage and ensure data avail- Moreover, we provide the content-addressed techniques
ability along with reliability using IPFS and blockchain. In which creates permanent storage while access of transaction.
this paper, we have considered the blockchain transaction as The provided hash by the IPFS point the same transaction
a image File. exactly owing to the content-addressed technique. Moreover,
AIM: We propose a blockchain data storage model that the content-addressed has many implication from a access
illustrates how to use IPFS networks to reduce the storage perspective like transaction storage, and transaction access
of transaction size of block in blockchain. Moreover, we have on IPFS. Once the hash of transaction gets created then it
also used the content-addressed based access of transactions. can be accessed by other peers in the network efficiently
Organization:This paper is organized as follows Section-II without revealing of location of the transactions [3].
describe related work of blockchain in healthcare, Section-III
describe the proposed structure, Section-IV describe the Im- Motivation: - The work in this paper is motivated by the
plementation and Result of the proposed framework, Section- following observation from the Literature [10], [12], [13], we
V concludes the paper. have notice that storage is a challenge in blockchain network
owing to the append only feature [4]. In this paper, we have
II. Related Work proposed IPFS-based blockchain storage system which elimi-
The authors in [7]–[9] has reported that currently a com- nates the limitations of existing blockchain storage. We have
plete ledger of Bitcoin network occupies up to 200GB, and also proposed the content-addressed access of transaction
this number constantly increasing at about 0.1 GB per day, rather than simply location based access of transactions.
causing great pressure on data storage. Moreover, the work in
III. Proposed Structure
[8] proposed network coding-based distributed storage (NC-
DS) framework, which divides the transactions in different A. Work flow synchronization of Miners
block and again block into sub-block.The each sub-block is There are two major work flow is performed by miners
encoded and distributed to all the peers(nodes) in network. i.e, participating in network process and mining process.
This scheme reduced the transaction size into some extends Process of new peers synchronizations:
but at the same time it also increased the complexity using The newly peer(node) join the network must have
encoding and decoding process. synchronized with the existing distributed structure. The
The author in [10] adopted the method of recording block transaction made by the peer must be synchronized to
techniques which does not record the expired transaction maintain the consistency in network. To maintain the
history. The header information of expired transaction is synchronization in network all the newly join peers must
stored in the recording block technique. This approach re- have to follow the consensus of blockchain.
duces the size of block in blockchain but at the same time
traceability of the complete transaction is not possible owing Process of Mining
to the header part storage only. The mining is process in which all the transactions is
The authors in [12] adopted the summary block techniques verified before submissions to the block of blockchain by
which includes the active transaction records in blockchain. miners. The verification of transaction involves format of
These records are kept in external file and the remaining old transactions, input of transactions, and the transactions meet
transactions are deleted from the blockchain. This technique the requirement of consensus. once the transactions are
reduces the size of block in blockchain but the drawback verified then it gets deposited to the mining pool. To add the
of this technique is that it can not retain the history of transactions in a new block, the miners verify the previous
blockchain. block. Furthermore, the previous block hash is added to the
The authors in [13] has discussed the account-tree techniques new block before packed into a new block in blockchain. To
which only includes transactions of the peers who is part of successful addition of a new block, the miners solve a crypto-
the networks. In this approach the account of each peers are puzzle by calculating the block hash [15]. The miners who
recorded in off-chain(Database)storage and matched during first calculates the block hash disseminates the block to the
the transactions. However, this approach does not maintain other peers(nodes) in a blockchain network. The other nodes
record of old transactions. the following approach is suitable in the network mutually verify the correctness of block.
to reduce the size of block owing to the does not include Moreover, the verified block by the peers get added by the
the history of old transactions. The account-tree techniques other miners as a new block in blockchain network [16].
does not suitable for complete traceability of blockchain. As shown in Algorithm 1, the files are uploaded by
The above techniques has some limitations and drawback, the peers of the network. to store the files into the IPFS
and the size of the transactions is also not reduced effectively. distributed file storage ipfsapi has been used. The connection
In this paper, we are using blockchain and IPFS to manage of ipfs is established on the port number 5001. The hash
the block transactions. We are using image file as block value (Hv) is calculated in binary format and then hash of

247
 2019 Fifth International Conference on Image Information Processing (ICIIP)

Algorithm 1: Algorithm for content addressed hash

Input: Image Files
Output: Content-addressed hash of the File
//the file uploaded by the peers in the network
f ile ← request.f iles[0 upload0 ]
//initiated the IPFS distributed file storage
api ← ipf sapi.connect(0 localhost0 , 5001)
//original file addition into IPFS storage
res ← api.add(f ile)
// file to binary conversion
Hv ← convert(f ile, binary)
// created the hash message digest of the file (create
hash − > ch)
digest ← ch(0 sha2560 ).update(f ile).digest()
// created the message digest using salting methods
M ds ← (digest.bytelength.toString(16),0 hex0 )
// combining binary conversion and salting message
digest for content-addressed hash
content addressed ← combine(Hv, M ds, digest)
// returning content addressed hash to store in Fig. 1. Working Model of Blockchain and IPFS
blockchain
Return content addressed hash
C. Implication of Content-addressed Model of IPFS
1) To increases the integrity of Transaction: The IPFS
the file gets created using sha-256 method [11], [17]. The content-addressed increases the integrity of transaction ow-
salting method approach is applied to convert the message ing to the fingerprint(hash). The obtained hash from IPFS
digest into the hex format. Finally, the message digest (Mds) makes transaction reliable and resilient. This types of finger-
and hash value are combined together to generate content- print validation becomes difficult with location-addressed.
addressed hash of the file. The content-addressed hash is 2) Easy to Access Transaction: In the IPFS distributed
stored into the blockchain network in order to reduce the network, as long as one node has copy of the transaction
size of the distributed ledger. then other peers will be able to access it. The attractive
feature of the IPFS is the hash of the transaction get
B. IPFS based blockchain storage data model changed when transaction get updated without changing
the peers connected to the IPFS hash. The content-addressed
Introduction to IPFS techniques of IPFS signify the importance of transaction and
IPFS is a distributed data storage which provides content provides belief to the peers that they are accessing exact
addressed and allocates the unique hash for stored file. IPFS content which was originally stored.
facilitate the high-throughput and efficient storage model 3) Easy to Distribute Transaction: In the IPFS peers can
with concurrent access. The IPFS creates hash which is access the content and distribute the content to others peers.
46 bytes long. Therefore, storing transaction data in IPFS The ”PIN” command is used to distribute the content to oth-
and storing the hash returned by IPFS into the block of ers peer. To get the details of ”PIN” command must visit the
blockchain provides large reduction in storage space. URL ”https : //docs.ipf s.io/ref erence/api/cli/#ipf s −
Blochchain Storage Model pin”
To provide the efficient blockchain storage model and the
feature of IPFS, the IPFS-based data storage model has been IV. Implementation and Results
proposed for blockchain, as shown in Fig.1.
In this model the miners collect the transactions and The implementation of our proposed IPFS-based
puts the valid transactions into the mining pools and stores blockchain storage model is carried out in the IPFS
them into the IPFS distributed network. The IPFS provides distributed file sharing system where each transaction
the unique hash for the stored transactions during the stored on IPFS and calculated unique hash is stored in
mining process which is use to create a new block. The blockchain transaction. The experimental setup consists of
stored transaction can be accessed by the unique hash value python anaconda, python flask. The setup is performed on
provided by the IPFS which is known as content addressable Intel(R) Xeon(R) W-2175 CPU @ 2.50GHZ running Window
access. The proposed technique does not disclose the actual x64- based processor with 128 GB of RAM and 1 TB of local
access name of the transaction. storage.

248
 2019 Fifth International Conference on Image Information Processing (ICIIP)

A. File upload to IPFS

As shown in Fig.2 and Fig.3, the transaction(Image file) is
uploaded onto the IPFS. The transaction which is large in
size should not be stored in the block of blockchain directly.
To achieve the efficient storage of the transaction we are
storing the transaction on IPFS.

Fig. 4. To retrieve the unique hash of the transaction

Fig. 2. Image as transaction on Blockchain using IPFS

Fig. 5. To retrieve the unique hash of the transaction

contains any IPFS hash value. Every block contains times-

Fig. 3. Image as transaction on Blockchain using IPFS
tamp which denotes the creation time of block to ensure the
reliability. The prrof-of-work maintains consistency of the
B. Process of Mining to retrieve the IPFS Hash block over the peers of blockchain. AS shown in Fig.6, the
length denotes the chain size of blockchain. The resultant
As discussed in the proposed structure, the transaction is
blockchain facilitate availability, reliability and immutability
stored to the IPFS. The IPFS provides unique hash of the
and reduced size of the transactions.
transaction during mining process as shown in Fig.4 and
Fig.5. We have mined the both of the transaction performed
in the Fig2. and Fig.3, and retrieved the unique hash for
both the transaction along with their size. The mining
process includes the index number which is block number
in blokchain and the previous hash maintain the chain of
block. Moreover, the list of transactions can be mined during
the mining process to create a single block in blockchain.
The generated IPFS hash value is 46 bytes which provides
efficient and less storage size than original size of the file
i.e., 4KB as shown in Fig.5. In the mining process we are
generating the IPFS hash value rather than to store the
original size of the transactions.

C. List of transactions in blockchain

The list of transactions are shown in Fig.6, which contains
all the IPFS hash value with their block number(index
number). The first block is genesis block which does not Fig. 6. Blockchain Transactions List

249
 2019 Fifth International Conference on Image Information Processing (ICIIP)

D. Content-Addressed based access of transaction

As shown in Fig.7, the transaction is accessed by their
hash value which is provided by the IPFS at the time of
mining process. The IPFS follow content-addressed storage
rather than location-addressed. In location addressed the
storage keeps the record of list, or directory, of the loca-
tions. However, the location-addressed reveal the path of the
transaction when request is made for a particular transaction.
The content-addressed storage (IPFS) keeps record of hash
value of the transaction rather than keeping the record of
location of transaction. The content-addressed storage does
Fig. 8. Connection to IPFS distributed File storage system
not reveal the location of the transaction , hence keeping the
transaction on IPFS distributed storage is much efficient in-
contrast with location-addressed storage. As shown in Fig.7,
we are accessing the transaction using their content-address.
To see the transaction on IPFS using content-addressed

Fig. 9. Connection to IPFS distributed File storage system

utilized when transactions gets stored into IPFS and “OUT”

bandwidth is utilized during transactions accessed from IPFS.

V. Conclusion
In this paper, we have designed the IPFS-based storage
model for blockchain to mitigate the drawback of storage
Fig. 7. Transaction access based on the IPFS hash and transaction access of a block in blockchain network. The
result of our IPFS storage model provides efficient storage
access the url must be followed ”localhost:8080/ipfs/<hash space owing to the content-addressed hash of the trans-
of file >” actions. Moreover, the content-addressed scheme is applied
E. IPFS Connection to access the transactions of block using the hash value
provided by the IPFS distributed storage. In the proposed
Fig.8, shows the connection of IPFS to see the details of model, we are storing the hash of the transaction rather than
status, files and peers on distributed environment. The status storing the complete transaction or original transaction to
describes bandwidth occupied by the IPFS during storage ensure the efficient storage scheme for blockchain network.
and access of transaction. As shown in Fig.9 and Fig.10, the our proposed structure of storage model can be utilized
details of all peers can be verified in IPFS such as how many with the other types of transaction files such as video and
peers are connected at a time and how many are accessing audio on blockchain. The existing storage model like bitcoin,
the resource on IPFS. The all peers are verified by the ”PEER ethereum, hyperledger suffers from the storage of bulky data
ID” and their ipfs address. Moreover we can also see the in the distributed ledger of blockchain network. Thus, our
peers location on IPFS. model can also be applied on existing storage model to
To establish the connection with IPFS the url must be reduce the size of each block.
followed ”127.0.0.1:5001/webui”.
References
F. Access of Bandwidth by IPFS
[1] Andreas, M., and Masteing Bitcoin. ”Unlocking Digital Cryptocurren-
The process of data storage and access on IPFS distributed cies.” (2015): 49-68.
network takes bandwidth which is shown in Fig. 11, There [2] Zheng, Z., Xie, S., Dai, H., Chen, X., & Wang, H. (2017, June). An
overview of blockchain technology: Architecture, consensus, and fu-
are two different types of bandwidth is occupied by the IPFS ture trends. In 2017 IEEE International Congress on Big Data (BigData
such as “IN” and “OUT” bandwidth. The “IN” bandwidth is Congress) (pp. 557-564). IEEE

250
 2019 Fifth International Conference on Image Information Processing (ICIIP)

IPFS-Based Storage Model for Blockchain. In 2018 IEEE/WIC/ACM

International Conference on Web Intelligence (WI) (pp. 704-708). IEEE.
[14] Benet, J. (2014). Ipfs-content addressed, versioned, p2p file system.
arXiv preprint arXiv:1407.3561.
[15] Tosh, D. K., Shetty, S., Liang, X., Kamhoua, C. A., Kwiat, K. A., & Njilla,
L. (2017, May). Security implications of blockchain cloud with analysis
of block withholding attack. In Proceedings of the 17th IEEE/ACM
International Symposium on Cluster, Cloud and Grid Computing (pp.
458-467). IEEE Press.
[16] Hao, Y., Li, Y., Dong, X., Fang, L., & Chen, P. (2018, June). Performance
Analysis of Consensus Algorithm in Private Blockchain. In 2018 IEEE
Intelligent Vehicles Symposium (IV) (pp. 280-285). IEEE.
[17] Kumar, A., Ghrera, S. P., & Tyagi, V. (2015). Modified buyer seller wa-
termarking protocol based on discrete wavelet transform and principal
Fig. 10. Information of IPFS peers component analysis. Indian Journal of Science and Technology, 8(35),
1-9.

Fig. 11. Transaction submission and Access Bandwidth by IPFS

[3] Benet, J. (2014). Ipfs-content addressed, versioned, p2p file system.

arXiv preprint arXiv:1407.3561.
[4] Chen, Y., Li, H., Li, K., & Zhang, J. (2017, December). An improved
P2P file system scheme based on IPFS and Blockchain. In 2017 IEEE
International Conference on Big Data (Big Data) (pp. 2652-2657). IEEE.
[5] Napoli, C., Pappalardo, G., & Tramontana, E. (2014, June). Improving
files availability for bittorrent using a diffusion model. In 2014 IEEE
23rd International WETICE Conference (pp. 191-196). IEEE.
[6] Alessi, M., Camillo, A., Giangreco, E., Matera, M., Pino, S., & Storelli,
D. (2018, June). Make Users Own Their Data: A Decentralized Personal
Data Store Prototype Based on Ethereum and IPFS. In 2018 3rd Inter-
national Conference on Smart and Sustainable Technologies (SpliTech)
(pp. 1-7). IEEE.
[7] Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system.
[8] Dai, M., Zhang, S., Wang, H., & Jin, S. (2018). A low storage room
requirement framework for distributed ledger in blockchain. IEEE
Access, 6, 22970-22975.
[9] Clark, J. B. A. M. J., Edward, A. N. J. A. K., & Felten, W. (2015).
Research perspectives and challenges for bitcoin and cryptocurrencies.
url: https://eprint. iacr. org/2015/261. pdf.
[10] Dorri, A., Kanhere, S. S., & Jurdak, R. (2017, April). Towards an opti-
mized blockchain for IoT. In Proceedings of the second international
conference on Internet-of-Things design and implementation (pp. 173-
178). ACM.
[11] Kumar, A., Ghrera, S. P., & Tyagi, V. (2015). A comparison of buyer-
seller watermarking protocol (BSWP) based on discrete cosine trans-
form (DCT) and discrete wavelet transform (DWT). In Emerging ICT
for Bridging the Future-Proceedings of the 49th Annual Convention of
the Computer Society of India (CSI) Volume 1 (pp. 401-408). Springer,
Cham.
[12] Palai, A., Vora, M., & Shah, A. (2018, February). Empowering light
nodes in blockchains with block summarization. In 2018 9th IFIP
International Conference on New Technologies, Mobility and Security
(NTMS) (pp. 1-5). IEEE.
[13] Zheng, Q., Li, Y., Chen, P., & Dong, X. (2018, December). An Innovative

251