Content uploaded by William J Buchanan
Author content
All content in this area was uploaded by William J Buchanan on Aug 28, 2020
Content may be subject to copyright.
Content uploaded by William J Buchanan
Author content
All content in this area was uploaded by William J Buchanan on Aug 28, 2020
Content may be subject to copyright.
1
TRUSTD: Combat Fake Content using Blockchain
and Collective Signature Technologies
1st Zakwan Jaroucheh Blockpass ID Lab
Edinburgh Napier University
Edinburgh
z.jaroucheh@napier.ac.uk
2nd Mohamad Alissa Blockpass ID Lab
Edinburgh Napier University
Edinburgh
m.Alissa@napier.ac.uk
3rd William J Buchanan Blockpass ID Lab
Edinburgh Napier University
Edinburgh
w.buchanan@napier.ac.uk
Abstract—The growing trend of sharing news/contents,
through social media platforms and the World Wide Web has
been seen to impact our perception of the truth, altering our
views about politics, economics, relationships, needs and wants.
This is because of the growing spread of misinformation and
disinformation intentionally or unintentionally by individuals and
organizations. This trend has grave political, social, ethical, and
privacy implications for society due to 1) the rapid developments
in the field of Machine Learning (ML) and Deep Learning (DL)
algorithms in creating realistic-looking yet fake digital content
(such as text, images, and videos), 2) the ability to customize
the content feeds and to create a polarized so-called “filter-
bubbles” leveraging the availability of the big-data. Therefore,
there is an ethical need to combat the flow of fake content. This
paper attempts to resolves some of the aspects of this combat by
presenting a high-level overview of TRUSTD, a blockchain and
collective signature based ecosystem to help content creators in
getting their content backed by the community, and to help users
judge on the credibility and correctness of these contents.
I. INTRODUCTION
Fake news has become increasingly one of the main threats
to democracy, journalism, and freedom of expression. It has
weakened public trust in governments and its potential impact
on the contentious Brexit referendum and the equally divisive
2016 U.S. presidential election is yet to be realized [1]. Our
economies are also affected by the spread of fake news, with
fake news being connected to stock market fluctuations and
massive trades. For example, fake news claiming that Barack
Obama was injured in an explosion wiped out $130 billion in
stock value 1.
The news is a newly received or noteworthy information,
especially about recent not-necessarily political events. De-
ception can be described as an act of "intentionally causing
another person to have or continue to have a false belief that is
truly believed to be false by the person intentionally causing
1https://www.forbes.com/sites/kenrapoza/2017/02/26/can-fake-news-
impact-the-stock-market
the false belief by bringing about evidence on the basis of
which the other person has or continues to have that false
belief" [2]. With the ease of sharing information on social
media platforms, the rapid accessibility of uploaded content
on the World Wide Web, and the rapidly progressing fields
of Artificial Intelligence (AI) and Machine Learning (ML),
deception has far more potential in altering an individual’s
perception and influencing their decisions and choices than it
has ever had. In this context, various issues can be identified:
1 - Fake content can have a strong impact on listeners
and readers: Experts distinguish between misinformation and
disinformation [3]. Misinformation is a false or misleading
piece of information —e.g., sharing a fraudulent content with-
out verifying the authenticity of its source is misinformation.
Disinformation is deliberately falsified information to obscure
the truth —e.g., spreading false content with the intention
to harm an individual’s reputation. In this paper we refer to
the misinformation, disinformation and artificially generated
content as deceptive artifacts or fake content. Fake content
is a general term which would include the term fake news.
Throughout the paper, we will use the terms fake news and
fake content interchangeably. Repeated exposure to a piece of
information makes it familiar, until it is eventually perceived
as acceptable and valid. This phenomenon is known as the
"illusory truth" effect [2].
2- Individuals are victims of the "filter bubbles": Social
media websites, including Facebook, Twitter and LinkedIn are
said to construct “filter bubbles” that allows users to only view
content they agree with or that aligns with their pre-existing
beliefs [4]. A filter bubble is an individual’s personal and
unique online space, the nature of which depends on their
online identity (e.g. preferences, behaviour, believes, etc.).
Individuals may not have full control on the construction of
this identity. In addition, individuals have no say regarding
what penetrates into the space and what gets filtered out.
Therefore, individuals are now victims of the algorithms that
2
these centralised platforms use to filter and sort the content
(content feed). There is no guarantee that these algorithms
are not designed to favor one content over another in order
to maximise the revenue or achieve a specific objective (e.g.
Cambridge Analytica scandal).
3- Lack of transparency and traceability: Anyone has the
freedom to publish any content on any social media platform
without any scrutiny about the correctness or the credibility of
the content. Individuals do not have any mechanism to trace
back the originator of the content. In addition, if the content
is published in a news agency Website, individuals do not
have any clue who has approved that content and what is the
approval process.
4- Lack of empowering tools: There is a debate whether
the centralised social media companies should intervene in
controlling the political adverts on their platforms or not.
Some people argue in favour of banning these adverts, while
others argue that these private companies should not make
the decision on behalf of the user and these adverts should
be shown to the user, and it is up to the user to make
their mind. We have two issues with this argument: 1) since
it is an advertisement platform, their algorithm could be
designed to allow these adverts to reach to the users in a way
that maximises the profit and maybe maximises the intended
impact on the users, 2) there is no helpful tools for the users
to assess the credibility and the correctness of the content.
The users are making their own judgement on any content in
an ad-hoc manner. Verifying the authenticity and credibility
of every content is a daunting task for the users. That is why
some users prefer to trust a few sources of news/information.
In order to remedy to the above issues we propose TRUSTD,
an ecosystem that allows content readers (users) to assess the
correctness and credibility of the content they receive. We
believe that there is a need to empower the user with a tool
that allows them to delegate the task of verifying the content
to a set of trusted parties of their own choice. That way, users
will be able to assess the credibility of the content themselves
based on their level of trustiness of each of their trusted
parties. In addition, since this is a global issue, TRUSTD is
a decentralised and open platform where everyone is able to
join the system.
Today, Distributed Ledger Technologies (DLTs) and specifi-
cally blockchain, the Decentralised Identifiers (DIDs) [5], and
collective signature [6] present opportunities for stakeholders
and policymakers as potential technologies that can help to
combat digital deception. These technologies enable security
and trust in a decentralised Peer-to-Peer (P2P) network without
any central managing authority. There are only a few articles
of the literature that use blockchain to combat digital deception
and they are mostly focused on tracing the source of the
information. To the knowledge of the authors, this is the first
article that proposes a user-centred approach to empower the
user with the a tool to identify digital deception using the
above technologies. In this paper, we are focusing on the
human element when addressing the fake content issue. In
other words, it is true that the machines are able to detect
the fake content to some extent but there is no replacement to
the human intervention. In addition, TRUSTD does not inhibit
the current existing automated tools to detect the fake content,
these deep tools can be included as part of the user policy
rather than rely only on the quality of these sophisticated tools.
The rest of this paper is organized as follows. Section II
reviews the state-of-the-art of current digital deception and
the involved technologies and approaches used to combat it.
Section III provides a background on the Schnorr signing and
how it is used in the collective signing algorithm. In Section
IV, we list the main driving requirements that drive the design
of TRUSTD. Section V describes the details of the proposed
architecture and its implementation. Finally, Section VI is
devoted to conclusions and future work.
II. RE LATE D WORK
Recently, Wikipedia co-founder Jimmy Wales has launched
"WikiTribune"2, a platform to combat the low-quality con-
tent. This platform is designed for small, niche communities
that can sustain themselves where almost everything on the
platform is editable (similar to Wikipedia) and the users are
responsible to correct the contents even those written by other
users. However, this might not be an effective approach to
combat the fake content as this platform can be turned into a
source of fake news.
Governments worldwide are taking various measures to
prevent the widespread of fake news, which may be driven
by different motivations and which may undermine national
security [7]. These measures include the introduction of new
legislation e.g. 1) new laws that would give governments more
powers to hold technology companies (e.g., Facebook, Twitter
and Google) and individuals accountable for the spread of
fake news, and 2) new laws that would seek to counter the
impact of automated social media accounts (bots). Although
these laws are important and in the right direction, they hold
the technology companies accountable of identifying the fake
news on behalf of the content readers (users). In this respect,
users have to trust that these companies implement adequate
algorithms to remove the fake news as quickly as possible.
However, this is not enough because the user has to trust
the criteria followed by these algorithms to identify the fake
news. The user should be empowered by tools to allow them to
specify their trust policy in order to determine the credibility
of the content.
In response, technology companies are defending them-
selves and are introducing mechanisms to detect and remove
fake news. For example, Facebook Journalism Project aims to
collaborate with content organizations and journalism experts
to improve the quality of information shared on the platform
[8]. Facebook tries to provide an improved ranking system for
posts shown in the News Feed and expediting the reporting
of misleading content. Google has presented a white paper
outlining measures to prevent the spread of disinformation
through their products [9]. The objective is to enhance its
search result ranking system by ranking news on the basis of
expertise, credibility and authority. Although these initiatives
are necessary, they still do not position the user in the center.
The user should be able to make their own judgement on the
2https://wt.social/
3
content based on their own criteria. We believe the fake news
problem cannot be solved by automated tools only; and these
tools should be complemented by manual verification process
conducted by humans.
The IEEE Global Initiative has published a report titled
“Ethically Aligned Design” [10] on guiding and encouraging
the development of autonomous systems that are primarily
focused on human well-being and protecting human rights
through preventing misuse of AI, ensuring system transparency
and developing a framework for developer accountability. The
goal of The IEEE Global Initiative is that Ethically Aligned
Design will provide pragmatic and directional insights and
recommendations, serving as a key reference for the work
of technologists, educators and policymakers in the coming
years. TRUSTD is inspired by the general principals of the
above initiative such as human rights, well-being, data agency,
transparency and accountability; and it uses these principals as
a guideline to derive the driving requirements in the design of
an open and decentralised ecosystem to combat the flow of
fake content.
The authors in [1], presented four perspectives to study fake
content: knowledge-based, style-based, propagation-based and
credibility based. In knowledge-based approach, one aims to
analyze and/or detect fake news, using a process known as
fact-checking. Manual fact-checking can be divided into (I)
expert-based and (II) crowd-sourced fact-checking. Expert-
based fact-checking relies on domain-experts to verify the
given news contents, therefore, it leads to highly accurate
results. Recently, many websites have emerged to provide
expert-based fact-checking services. For example, PolitiFact
provides “the PolitiFact scorecard”, which presents statistics
on the authenticity distribution of all the statements related
to a specific topic. Another example is HoaxSlayer, which
classifies the articles and messages into e.g., hoaxes, spams
and fake news. A comprehensive list of fact-checking websites
is provided by Reporters Lab at Duke University3, where
over two hundred fact-checking websites across countries and
languages are listed. Although these websites can provide
ground-truth for the detection of fake content, these websites
are still silo-ed and centralised and they have their own
"experts" and their own methodology of identifying the fake
news.
Crowd-sourced fact-checking relies on a large number of
fact-checking individuals and therefore, it is less credible and
accurate due to the political bias of these individuals and
their conflicting annotations. Hence, one often needs to (i)
filter non-credible individuals and (ii) resolve conflicting fact-
checking results. An example is Fiskkit4, where users can
upload articles, and provide ratings and tags for sentences
within articles. TRUSTD can be seen as a crowd-sourced fact-
checking approach but it is the content creator who chooses the
fact-checkers and it is the user who determines the content’s
credibility based on their trust level in these fact-checkers.
Automatic fact-checking techniques have been developed,
heavily relying on Information Retrieval (IR) and Natural
3https://reporterslab.org/fact-checking/
4https://fiskkit.com/
Language Processing (NLP) techniques. The overall automatic
fact-checking process can be divided into two stages: (I) fact
extraction (also known as knowledge-base construction) and
(II) fact-checking (also known as knowledge comparison) [1].
Knowledge-based approaches aim to evaluate the authenticity
of the given content, while style-based approaches aim to
assess the intention behind publishing the content. Credibility-
based approaches evaluate the content based on content-
related and social-related information. The AI Foundation has
developed an intelligent software called “Reality Defender”, to
detect potentially fake media in the digital world5. This soft-
ware runs AI-driven analysis techniques to detect alterations
in the scanned images, videos and other media, and allows for
reporting suspected fakes. In TRUSTD, these deep tools can
be included as part of the user’s trust policy, as will be seen
later, rather than rely solely on the quality of these tools.
Blockchain technology has been already leveraged to con-
tribute solutions to address the fake news problem. For exam-
ple, Shang et al. [11] trace the source of news by keeping
a ledger of timestamps and the connections between the
different blocks. Huckle et al. [12] introduced a blockchain-
based application that is capable of indicating the authenticity
of digital media. Using the trust mechanisms of blockchain
technology, the tool can show the provenance of any source of
digital media, including images used out of context in attempts
to mislead. However, the authors mentioned that although
their application has the potential to be able to verify the
originality of media resources, that technology is only capable
of providing a partial solution to fake content. We agree with
the authors that this is because the blockchain technology is
incapable of proving the authenticity of a content story and
that requires human skills.
III. BACKGRO UN D
A. Schnorr signing
In Schnorr signing, we can aggregate public keys of P
participants into a single signing key [13], and uses the
non-interactive version of the Fiat-Shamir heuristic 6. Using
Elliptic Curve methods, to sign a message we take a random
value (k) and a private key value (d) and a generator point G
(Gis a base point on an elliptic curve) and compute:
Q=dG (1)
and:
R=kG (2)
For it to be non-interactive we then calculate:
e=H(RkM)(3)
and then:
s=k−ed (4)
The signature is then (s, e). To verify we compute:
rv=sG +eQ (5)
5https://aifoundation.com/responsibility/
6https://en.wikipedia.org/wiki/Fiat-Shamir_heuristic
4
and then:
ev=H(rvkM)(6)
We then check that rv=ev.
Each participant has a private key (ai) and a public key
Ai=aiG. We can then determine the aggregate public key
with:
A=X
i∈P
Ai(7)
B. Collective Signing
Within legal infrastructures, we might have several wit-
nesses W, and we ask a number of the witnesses W0to verify
that something is correct. If one of the witnesses cannot verity
the information, we would highlight a problem. Let us say
we have a controller on a network, and a number of selected
trusted witnesses. Each of the witnesses can then check all of
the messages sent by the nodes on the network, and if one
of them determines a problem, they can tell the rest of the
network. In this respect, every message (M) is collectively
signed by Wwitnesses.
The collective signing (CoSi) algorithm is defined by Syta
et al [6]. With CoSi (collective signing), there are four phases
involving Pparticipants and where the leader has an index
value of zero. Each participant has a private key (ai) and a
public key (Ai=aiG, and where Gis a base point on an
elliptic curve). We then determine the aggregated public key
with [6]:
A=X
i∈P
Ai(8)
Announcement: Initially the leader broadcasts a message
(M) that it wants the participants to sign.
Commitment: Each node iwill pick a random scalar
(vi) and determines their commitment (Vi= [vi]G). Each
commitment is then sent to the leader, who will wait for a
specific amount of commitments (P0) to be received. The
leader then creates a participant bitmask and aggregates all
the received commitments:
V=X
j∈P0
Vj(9)
and creates a participation bitmask Z. The leader then
broadcasts Vand Zto the other participants.
Challenge: Each of the participants computes the collective
challenge (using the hash function H):
c=H(V||A||M)(10)
and send the following back to the leader:
ri=vi+c×ai(11)
Response: The leader will wait until the participants in P0
have sent their responses. Once received, the leader computes
the aggregated response:
r=X
j∈P0
rj(12)
and publishes the signature of the message (M) as:
(V, r, Z )(13)
Each node can then check their own signature value and
agree with the leader.
In order to handle large numbers of participants during
signature generation efficiently, CoSi protocol uses a tree-
shaped network communication overlay [14]. Any tree used
by CoSi should be a complete tree for performance reasons.
The leader is the root node of the tree and is responsible for
creating the tree. An intermediate node is a node who has
one parent node and at least one child node. A leaf node is
a node who has only one parent and no child nodes. The
leader multicasts a message to his direct child nodes. Upon
reception of a message, each node stores the message and
multicasts it further down to its children node, except if the
node is a leaf. Each node generates its response. Each leaf
node sends its response to their parent and is allowed to leave
the protocol. Each other node starts waiting for the responses
of its children. When the root node receives all the responses
from its children, it can generate the signature.
C. Decentralised Identifier (DID)
The self-sovereign identity (SSI) refers to an identity man-
agement system which allows individuals to fully own and
manage their digital identity [15]. The World Wide Web Con-
sortium (W3C) working group on verifiable claims states that
in a SSI system users exist independently from services [16].
This highlights the contrast to current identity management
which either relies on a number of large identity providers such
as Facebook and Google or the user has to create new digital
identities at each individual service provider. SSI is enabled
by the new development of blockchain technology. Through
the trustless, decentralised database that is provided by a
blockchain, classic Identity Management registration processes
can be replaced [15].
A Decentralised Identifier (DID) is comprised of a scheme
as well as a method and method specific identifier. The method
closely resembles the namespace component of an Uniform
Resource Name (URN) 7. Each distinct blockchain or rather
each identity registry constitutes its own namespace while
the blockchain specific identifiers specify the actual identity
addressed by the DID. An example for such a DID path would
be: did:examplechain:123456789. The DID document is the
key to the decentralised identity.
A DID is a unique identifier that can be resolved to a DID
document [17]. This document contains cryptographic material
and authentication suites enabling identification of the DID
subject. They also contain service endpoints allowing secure
encrypted communication to the DID subject. A public DID is
a DID that is registered on a public distributed ledger meaning
DID is resolvable to a document and hence verifiable by
7https://en.wikipedia.org/wiki/Uniform_Resource_Name
5
anyone. There should be some restrictions or process by which
DIDs can be registered to a public ledger. Once registered it
is possible to use that DID as a root of trust.
Establishing pairwise private DID connections can be done
by first contacting the endpoint (i.e. agent) resolved from
the public DID document. Sending a new unique DID and
DID document along with proof you are the DID subject. In
response you should receive a new unique DID signed by the
public key of the public DID. Together these two DIDs form
a private unlinkable connection between two parties complete
with private endpoints and the cryptographic material required
to verify the origin and integrity of the communication.
D. DLTs and blockchain capabilities
DLTs like Tangle or blockchain are able to provide seamless
authentication, efficient and secure data storage, robustness
against attacks, scalability, transparency and accountability.
Such features can play an effective role in combating fake
content, considering that transactions cannot be tampered
once they have been distributed, accepted and validated by
a network consensus and stored in blocks [18]. Moreover,
transactions are easily auditable by all the involved stakehold-
ers. More detailed information on how to design a blockchain
according to the business needs and deployment environment
can be found, for example, in [19] [20].
IV. DRIVING REQUIREMENTS
We believe that combating the fake content cannot be a pure
technical solution and that a human element should always
exist. In the proposed ecosystem, we can identify two types
of users:
- Individuals and organizations who produce the content.
We call them content creators (CCs). These creators may have
different agendas and they could be honest or dishonest in
reporting any news or spreading any information.
- Content readers: We call them users. These users are the
consumers of the content.
The following are the set of requirements that we believe
TRUSTD needs to meet:
A. Traceability of origin
The ecosystem should allow the originator of the content to
be identified at any point of time. That could be achieved
by storing the content along with the originator ID in an
immutable ledger that cannot be maliciously altered (i.e.
blockchain).
B. Traceability of approval
In TRUSTD, every CC can request other actors to "sign"
their content. This "signature" means that the signing actor
agrees on the correctness of the content. This is necessary
in order for the user to be able to assess the credibility of
the content, as will be seen in the next sections. Once any
actor provides their signature (witnessing the correctness of
the content), this signature should be stored in a way that
this actor cannot deny their action at any time. However, the
ecosystem should allow any actor to change their mind and
revoke their signature after, for example, they discover that
the content was incorrect. For transparency purposes, the list
of actions of any actor should be captured in the immutable
ledger.
C. Accountability
There is a need to ensure that the signing actors cannot deny
their involvement in signing off the content when the content
is discovered to be false.
D. System openness
The ecosystem should allow any individual or organization
to join and be part of the approval actors. For example, these
individuals can be journalists, editors, activists, etc. who have
different levels of expertise; and the organizations can be any
news agency. However, since the fake content concerns all
people and it affects everyone’s life, the system should be
open to allow anyone to be part of the signing actors - we
call them appraiser actors (AAs). This is especially important
for the content that is originated from people (non-journalists)
such as abnormal accidents or natural disasters, and that the
only witnesses are the people seen at the accident location.
E. The content should be evaluated by the user
It is true that the ecosystem allows the invited AAs to sign
off the contents they receive from CCs. However, the system
should keep the user in the center and it should empower them
with a mechanism to judge on the credibility of the content.
For that objective, the user should be able to select the list
of AAs and assign them a trust level value which is in the
range [0, T ].Tis the maximum trustiness value. For example,
a trust level of Tmeans the user completely trusts that actor.
A trust level of 0 means the user does not trust the actor.
The user can determines the credibility of the content by
calculating the following formula:
C=X
i∈N
Ti/(T∗N)(14)
T i: The trustiness value of the actor i.N: The length of
the actor list added by the user. The result is a number in the
range [0,1]. This number indicates the level of credibility of
the content.
F. The content should be backed by actors
Before publishing their content to the social media and news
websites, the CCs can get an "approval" from a set of AAs of
their choice. These AAs can be selected by the CCs themselves
or by the system in automated way based on a specific criteria.
For example, the system can extract the main keywords of
the content and send it to all AAs that are tagged with these
keywords. In addition, these AAs can specify a policy (using
the DID system) to accept signature requests only from CCs
who can prove that they have specific attributes, for example,
holding a specific degree. These attributes should be signed
by an entity trusted by the AA.
6
The approval comes in a form of a collection of signatures
from these AAs indicating that they agree on the correctness
of the content. We can imagine here different signature types.
For example, the CC may obtain a collection of signatures
from the AAs who share the CC’s opinion, or who support
the CC’s call. In this paper, we focus on the signature that
means that the AA agree on the correctness of the content.
G. The ecosystem should be decentralised and scalable
Since the fake news/content is a global issue, the system
should not be centralised. Instead, the system should provide a
protocol between the different users and actors which are reg-
istered in different subsystems owned by different companies
or organisations. However, these subsystems should have one
common denominator: the identifier of the users and actors.
For that aim, the decentralised identity (DID) [5] would be a
perfect fit in this context. In this respect, the actors can create
their own DIDs and publish them on their websites or their
social media profiles. That DID will be used for signature
verification as will be seen later.
V. SY ST EM DESIGN
Three main entities can be identified: 1) The content creator
(CC), 2) the appraisal actor (AA) who signs off any content,
and 3) the user who needs to assess the credibility of any
content. Each entity should be represented by a DID agent.
These agents should follow a specific protocol in order to
get any content signed off by the relevant AAs. The users’
agents should be able to verify any content based on the trust
policy specified by the user. Here is the proposed protocol as
illustrated in Fig 1:
1- An AA creates its own DID and publishes it somewhere
so that the CCs, users and other AAs know about it.
2- A CC creates the content in a form of a text, photo,
video, sound, or any other digital format.
3- Depending on the content scope, the CC can choose a
list of AAs that they are candidate to approve and sign off the
content. Alternatively the system recommends specific AAs.
The result is a set of DIDs which can be used to retrieve their
corresponding public keys.
4- A collective signature is generated in five steps over two
round trips between the agent of the CC and the agents of the
selected AAs, as follows:
AAs need not coordinate for the creation of their key-
pairs beyond selecting a common elliptic curve, and verifiers
can apply flexible acceptance policies beyond simple t-of-n
thresholds.
We use the following notation:
B: Generator of the group of AAs.
L: Order of the group generated by B.
(ai,Ai): Each AAigenerates their long term private-public
key pair (ai,Ai) as in EdDSA.
A: collective public key Agenerated from the public keys
of AAs.
N: denotes the list of AAs, the size of Nis denoted by n.
-Announcement Step: The CC broadcasts an announce-
ment message to the AAs indicating the start of a signing
process. This message contains the content itself.
-Commitment Step: Upon the receipt of the announcement
message, each AAigenerates a random secret riby hashing
32 bytes of cryptographically secure random data. Each ri
must be re-generated until it is different from (0mod L) or
(1mod L). Each AA then constructs the commitment Rias
the encoding of [ri]B, sends Rito the CC and stores the
generated rifor usage in the response phase.
-Challenge Step: The CC waits to receive the commitments
Rifrom the other AAs for a certain configurable time frame.
After the timeout, the CC constructs the subset Mof AAs
from whom he has received a commitment Ri and computes
the sum:
R=X
i∈M
Ri(15)
The CC then computes
c=SH A512(R||A||M)mod L (16)
The CC broadcasts cto all AAs.
-Response Step: Upon reception of c, each AA generates
their response:
si= (ri+c∗ai)mod L (17)
and send it to the CC.
-Signature Generation Step: The CC waits to receive
the responses sifrom the AAs for a certain configurable
time frame. After the timeout, the CC checks if he received
responses from all AAs in Mand if not he must abort the
protocol. The CC then computes the aggregate response
s=X
i∈M
simod L (18)
and initializes a bitmask Zof size nto all zero. For each
AAiwho is present in Nbut not in Mthe CC sets the i−th
bit of Zto 1, i.e., Z[i] = 1. The CC then forms the signature
sig as the concatenation of the byte-encoded point R, the byte-
encoded scalar s, and the bitmask Z. The resulting signature
is of the form:
sig =R||s||Z(19)
5- The CC sends a request to the underlying blockchain to
store the hash of the content along with the collective signature
in the blockchain.
6- In order to verify the credibility of a specific content,
the user agent calculates the hash of the content and uses it
as an input parameter to query the blockchain and to retrieve
the corresponding collective signature (if any).
7- As aforementioned, each user should be able to choose
a set of their trusted AAs. Each user agent maintains a list
of the public keys of those actors along with their trust level
values. Based on the retrieved collective signature, the user
agent obtains the public keys of the AAs who signed off the
content and then calculates the credibility value of the content
as described in section IV.E.
7
Fig. 1. The TRUSTD approach
VI. IMPLEMENTATION
An initial prototype has been built (Fig. 2 to 4). We use
Django version 2.2.7 to build a TRUSTD-based blog where the
users can register to the system, create contents, specify the list
of their trusted parties, assign a trust level to each actor in that
trusted list, send the contents to a subset of their trusted parties
list (Fig 2), view the names of the signed parties along with
the trustworthiness level of the related content (Fig 4), and
receive requests from other users to sign contents (Fig 3). We
use GO programming language version 1.13.4 to implement
CoSi8.
VII. DISCUSSION
A. Limitation of the TRUSTD approach
In its current implementation, TRUSTD has some limita-
tions that can be solved in future work:
- The authors in [21] introduced mBCJ, a secure two-
round multi-signature scheme. Their results show that mBCJ
is only marginally less efficient than CoSi, so that any protocol
based on the CoSi scheme should instead be built on the
provably more secure mBCJ scheme. The TRUSTD approach
is agnostic to the used collective signature scheme. We plan to
adopt the mBCJ scheme as a mechanism to collectively sign
any digital content.
- In some extreme cases, the content could be signed by one
AA who is not trusted by the user. For example, if the content
is a video of a publicly known person and this person signs
off this video as correct, but this person is not trusted by the
user, in our current implementation, the system recommends
to the user that the video is not credible even though that AA
is the only AA who can sign off this video. Therefore, we
need to extend the way we calculate the credibility of any
8https://asecuritysite.com/encryption/go_cosi
content to consider different dimensions, each of which takes
two extremes, for example: [liar, truthful], [enemy, friend],
[dishonest, honest], etc.
- In order for the CC to get their content signed off by the
AAs, they may need to wait for a certain time delay. This delay
is caused by two factors: 1) the round-trip time required to get
the commitments and signatures, and 2) the lack of incentive
for the AAs to sign off the content quickly. However this risk
is mitigated by the fact that the CC can initiate the sign off
request at any point of time, even after the content has been
published. In addition, they can publish their content (along
with the signatures they received so far), and then they can
initiate another request to get more signatures on the content.
- Once the collective signature along with the content hash
has been added in the blockchain, the AAs cannot revoke
their decision because there is no way to revert that addition
in blockchain. However, we are planning to implement a
mechanism to allow any AA to retrigger the sign off process
which will result in a new collective signature.
B. Advantages of the TRUSTD approach
TRUSTD approach can be applied not only for combating
the fake news, but it could be extended to other domains
as well. For example, a new type of a decentralised social
network can be built in which CCs and AAs help each other.
Contrast to the existing centralised social media networks,
a decentralised network can be built where CCs produce
different types of content and get different types of appraisal
such as "Agree" (I agree with the point of view of the content),
"Correct" (the content is correct), "Support" (I support the
call of the content), etc. (in form of signature) from differ-
ent individuals and organisations without any central entity
governing the whole network. For example, authors can get
their research conference/journal papers marked as "correct"
8
Fig. 2. The user assigns a trust level to each trusted party, and sends the
content to a subset of them
Fig. 3. The user receives requests from other users to sign contents
Fig. 4. The user can see the names of parties who signed the content, along
with the trustworthiness level of the content
by the reviewers, bloggers can get their opinion blogs marked
as "agreed upon" by a set of authors/people, petition creators
can get their petitions to request some change marked as
"supported" by a set of people, etc. In all cases, users are
able to judge on the credibility of any content based on their
trust policy.
VIII. CONCLUSION
We look at spreading the fake news as a special case of
spreading all kinds of deceptive contents. The use of online
social media to connect with people around the world is
rising sharply and it has become an important source for
the distribution of digital deception. This setting shifts the
responsibility of verifying the content to these centralised
social media platforms. In addition, the user has no say or
control over the filtering mechanism implemented by these
platforms. TRUSTD allows users to specifies their trusted
parties and to determine the credibility level of any published
content. We believe this is a step in the right direction where
the ecosystem is decentralised, all users, AAs and CCs own
their identities, and the user is in the center of the equation.
REFERENCES
[1] X. Zhou and R. Zafarani, “Fake news: A survey of research, detection
methods, and opportunities,” vol. 1, no. 1, 2018.
[2] J. Mahon, “Kant on lies, candour and reticence,” Kantian Review, vol. 7,
03 2003.
[3] T. Zubair, A. Raquib, and J. Qadir, “On combating fake news, misinfor-
mation, and machine learning generated fakes: Insights from the islamic
ethical tradition,” 10 2019.
[4] D. Difranzo and K. Gloria-Garcia, “Filter bubbles and fake news,”
XRDS: Crossroads, The ACM Magazine for Students, vol. 23, pp. 32–35,
04 2017.
[5] “Decentralized identifiers (dids) v1.0, w3c working draft,” 2019.
[Online]. Available: https://www.w3.org/TR/did-core/
[6] E. Syta, I. Tamas, D. Visher, D. I. Wolinsky, P. Jovanovic, L. Gasser,
N. Gailly, I. Khoffi, and B. Ford, “Keeping authorities" honest or bust"
with decentralized witness cosigning,” in 2016 IEEE Symposium on
Security and Privacy (SP). Ieee, 2016, pp. 526–545.
[7] G. Haciyakupoglu, J. Y. Hui, V. Suguna, D. Leong, and M. F. B. A.
Rahman, “Countering fake news: A survey of recent global initiatives,”
2018.
[8] A. Guess, J. Nagler, and J. Tucker, “Less than you think: Prevalence and
predictors of fake news dissemination on facebook,” Science Advances,
vol. 5, p. eaau4586, 01 2019.
[9] “How google fights disinformation,” 2019.
[10] “Ieee global initiative et al., ethically aligned design, ieee standards
v1,” 2016. [Online]. Available: https://standards.ieee.org/content/dam/
ieee-standards/standards/web/documents/other/ead_v2.pdf
[11] W. Shang, M. Liu, W. Lin, and M. Jia, “Tracing the source of news
based on blockchain,” 06 2018, pp. 377–381.
[12] S. Huckle and M. White, “Fake news: A technological approach to
proving the origins of content, using blockchains,” Big Data, vol. 5,
pp. 356–371, 12 2017.
[13] K. Ohta and T. Okamoto, “Multi-signature schemes secure against active
insider attacks,” IEICE Transactions on Fundamentals of Electronics,
Communications and Computer Sciences, vol. 82, no. 1, pp. 21–31,
1999.
[14] B. Ford, N. Gailly, L. Gasser, and P. Jovanovic, “Collective edwards-
curve digital signature algorithm,” Internet-Draft draft-ford-cfrg-cosi-00.
txt. IETF Secretariat, Tech. Rep., 2017.
[15] A. Mühle, A. Grüner, T. Gayvoronskaya, and C. Meinel, “A survey on
essential components of a self-sovereign identity,” 07 2018.
[16] “Verifiable claims working group frequently asked questions,”
2019. [Online]. Available: https://w3c.github.io/webpayments-ig/VCTF/
charter/faq.html
[17] “Decentralized Identifiers (DIDs) v0.11.” [Online]. Available: https:
//w3c-ccg.github.io/did-spec/
9
[18] A. Shahaab, B. Lidgey, C. Hewage, and I. Khan, “Applicability and
appropriateness of distributed ledgers consensus protocols in public and
private sectors: A systematic review,” IEEE Access, vol. PP, pp. 1–1, 03
2019.
[19] P. Fraga-Lamas and T. Fernández-Caramés, “A review on blockchain
technologies for an advanced and cyber-resilient automotive industry,”
IEEE Access, vol. 7, pp. 17 578–17 598, 01 2019.
[20] T. M. Fernández-Caramés and P. Fraga-Lamas, “A review on the
application of blockchain for the next generation of cybersecure industry
4.0 smart factories,” 2019.
[21] M. Drijvers, K. Edalatnejad, B. Ford, E. Kiltz, J. Loss, G. Neven, and
I. Stepanovs, “On the security of two-round multi-signatures,” 05 2019,
pp. 1084–1101.
