Dynamic and scalable multi-level trust management model for Social Internet of Things

Abstract

The Internet of Things (IoT) is a paradigm that has made everyday objects intelligent by offering them the ability to connect to the Internet and communicate. Integrating the social component into IoT gave rise to the Social Internet of Things (SIoT), which has helped overcome various issues such as heterogeneity and navigability. In this kind of environment, participants compete to offer a variety of attractive services. Nevertheless, some of them resort to malicious behaviour to spread poor-quality services. They perform so-called Trust-Attacks and break the basic functionality of the system. Trust management mechanisms aim to counter these attacks and provide the user with an estimate of the trust degree they can place in other users, thus ensuring reliable and qualified exchanges and interactions. Several works in literature have interfered with this problem and have proposed different Trust-Models. The majority tried to adapt and reapply Trust-Models designed for common social networks or peer-to-peer ones. That is, despite the similarities between these types of networks, SIoT ones present specific peculiarities. In SIoT, users, devices and services are collaborating. Devices entities can present constrained computing and storage capabilities, and their number can reach some millions. The resulting network is complex, constrained and highly dynamic, and the attacks-implications can be more significant. In this paper, we propose DSL-STM a new dynamic and scalable multi-level Trust-Model, specifically designed for SIoT environments. We propose multidimensional metrics to describe and SIoT entities behaviours. The latter are aggregated via a Machine Learning-based method, allowing classifying users, detecting attack types and countering them. Finally, a hybrid propagation method is suggested to spread trust values in the network, while minimizing resource consumption and preserving scalability and dynamism. Experimentation made on various simulated scenarios allows us to prove the resilience and performance of DSL-STM.

Appendix: Terminologies

Notation	Meaning	Definition	Other designations
Social Internet of Things
IoT	Internet of things	A paradigm where physical devices are embedded with sensors , software and other technologies to interact ad exchange data over the Internet	Connected environment
SIoT	Social Internet of Things	A paradigm born from the integration of the social component in the IoT	–
–	Device	Technical artefacts intended to furnish an interface between the digital world and the physical world	Things-objects-node-actor-entity
–	User	Humans participating in IoT interactions by providing or requesting services and assigning rates and recommendations	Human-participant-actor-node-entity-provider-requester-consumer-recommender
–	Service	An abstraction allowing the construction of complex software systems	Node-Entity
–	Provider	A user who provides services for other users through the devices he owns	–
–	Requester	A user who invokes services provided by other users	Consumer
–	Recommender	A user who assign a rate to a service he invoked	–
-	Resource	Computing or storing capability. Can also refers to devices or services made available for IoT users	–
-	Rating	A score given by a user to a service he has invoked	Rate-feedback-opinion-recommendation
–	Community	A set of user	–
Trust concept
–	Trust	A social and psychological concept involving two actors: a trustor and a trustee	Confidence-belief
–	Trustor	A user who decides to trust another	–
–	Trustee	A user who is trusted.	–
–	Intention	A social and psychological concept referring to aims or plans of a user	Good will
-	Ability	A term designing the possession of the means or skill to do something	Capability-competency
–	Expectation	A term that expresses the needs and the requirement of a user	–
–	Risk	A situation involving exposure to danger	–
-	Trust-dimension	Different aspects of the trust concept. In this work there is a trust-dimension for each kind of entity as the purpose of trust vary according the kind of entity	–
	Reputation system	Programs or algorithms that allow users to rate each other in online communities in order to build trust through reputation
Trust management process
TMM	Trust management mechanism	A mechanism which forms a layer superimposed on a given system and provides methods for assessing, propagating, storing and updating trust values for the system actors	–
–	Trust-features	Descriptors that make it possible to establish the criteria considered to compare and evaluate the network nodes	Indicator-factor-parameter-
TAM	Trust assessment model	The main component of a TMM designed for trust computation	–
–	Trust-composition	The first step of a TAM consisting in selecting trust features	Feature-engineering
–	Trust-aggregation	The second step of a TAM consisting in selecting a method to aggregate TrustFeatures on a trust value or a trust decision	–
–	Trust-propagation	The second phase of a TMM consisting on choosing a method for propagating trust-values computed by the TAM on the network	–
–	Trust-updating	The third phase of a TMM consisting on choosing a method for updating trust values	–
–	Node-ranking	A module consisting in providing scores to rank nodes according to many criteria	–
–	Attack-detection	A module aiming to classify nodes in malicious/legitimate and detect the type of launched attacks	–
–	Scalability	Consists of supporting a continuous and unlimited increase in the number of devices on the network	–
–	Dynamism	Consists of supporting a huge number of real-time interactions	–
-	Resources-efficiency	Consists of ensuring minimal consumption of computing and storage resources	–
Trust-Attacks modelling
TA	Trust-Attack	A set of actions performed by a malicious user in order to bias a reputation system	–
BMA	Bad-mouthing attack	A Trust-Attack where malicious nodes collude to destroy the reputation of a well-behaving one	–
BSA	Ballot-stuffing attack	A Trust-Attack where malicious disreputable nodes collaborate to promote their mutual reputations	–
SPA	Self-promoting attack	A Trust-Attack where a malicious node providing bad-quality service promotes his own reputation	–
DA	Discriminatory attack	A Trust-Attack where a malicious node attacks any other nodes without a strong relationship with him	–
OOA	On–off attack	A Trust-Attack where a malicious node provides bad services randomly instead of always providing the best services	–
OSA	Opportunistic-services attack	A Trust-Attack where a malicious node provides good services until he reaches a good reputation and then starts providing bad services	–
MRA	Malicious recommender attack	A category of Trust-Attacks where the node uses erroneous and unrepresentative ratings to skew the reputation system	–
MPA	Malicious provider attack	A category of Trust-Attacks where the node does not use rating but the chronological order of providing good/bad service to skew the reputation system	–
–	Malicious	A user who performed one or multiple Trust-Attacks	–
–	Legitimate	A user who does not launch Trust-Attacks	Benign-well-behaving
DSL-STM
Rep($u_i$)	Reputation	A Trust-Feature representing the global renown of a service provider in the network	–
Cred($u_i$)	Credibility	A trust-feature assessing how much the user’s rating match with his real opinion about an interaction he made	–
dExp($u_i,u_j$)	Direct experience	A trust feature representing the opinion of a user $u_i$ about his past interactions with another user $u_j$	–
RateF($u_i,u_j$)	Rating frequency	A trust feature allowing to estimate if a user $u_i$ is targeting another user $u_j$	–
Sim($u_i,u_j$)	Similarity	A trust feature consisting of measuring the degree of similarity between two users $u_i$ and $u_j$	–
RateT($u_i$)	Rating Trend	A trust feature used to understand the general behaviour of a recommender node in the network	–
relStren($u_i,u_j$)	Relationship Strength	A trust feature allowing to study the relationship between two users	–
Fluct($u_i$)	Fluctuation	A trust feature to depict the variation in the service quality for a provider node	–
DCC	Device capability criterion	A criterion permitting devices classification according to their storage, computing and communication capabilities	–
ELC	Energy limitation criterion	A criterion permitting to classify devices according to their power terminology and their strategy for using power	–
MSRC	Minimal security requirement criterion	A criterion permitting to classify devices according to their respect for minimal security requirements	–
$T_u(u_i)$	User trust score	A score estimating the degree of trust we can assign to a user based on his reputation	–
$T_d(d_j)$	Device trust score	A score estimating the degree of trust we can assign to a device	–
$T_s(s_k)$	Service trust score	A score estimating the degree of trust we can assign to a service according to its functional and no-functional characteristics	–
$fc(s_k)$	Functional characteristics	A device feature expressing if a service is able to achieve the task for each it was designed.	–
$nfc(s_k)$	Non-functional characteristics	Secondary characteristics that qualify a service such as its response time	–
$TN_i$	Trustful node	Nodes trusted because they do not provide services and are just designed to make trust calculations	–
$N_i$	Node	Nodes providing and consuming services	–
$f_h(x)$	Hash function	Any function used to map data of arbitrary size to a fixed size values	–
Experimental metrics			–
R	Recall	A performance metric computed as the fraction of relevant instances that were retrieved	–
P	Precision	A performance metric computed as the fraction of relevant instances among the retrieved instances. were retrieved	–
F–M	F-measure	A performance metric that combines recall and precision	–
NBQS	Number of bad quality services	A performance metric indicating the rate of propagation of poor-quality services	–