In this paper, we investigate the scope of a thing in a multiple IoT scenario. First we introduce the concept of scope in general, and we illustrate how it has been investigated and applied in social networking. Then, we de ne the scope of a thing in a Multi-IoT scenario, modeled as an extension of a Social Internetworking System, and we propose a formalization of scope allowing the computation of its values. Finally, we present a possible application of scope and describe some tests that we performed for evaluation purposes.
In the Concise Oxford Dictionary [
In the past, scope has been studied in the context of social networks. For
instance, in [
Parallel to the transition from social networking to social internetworking, in
the last few years, we are experiencing the presence of objects that are becoming
Copyright c 2019 for the individual papers by the papers authors. Copying
permitted for private and academic purposes. This volume is published and copyrighted by
its editors. SEBD 2019, June 16-19, 2019, Castiglione della Pescaia, Italy.
increasingly smart and social. This phenomenon is revolutionizing the Internet of
Things. As a proof of this, more and more authors are starting to study the
behavior of things, to talk about their pro les and their social interactions. In fact,
di erent architectures implementing these ideas are frequently proposed in the
current literature. Social Internet of Things (hereafter, SIoT [
In spite of this enormous interest that increasingly numerous researchers are
posing on the IoT, to the best of our knowledge, no analysis about the scope of
a thing in a Multi-IoT, or at least in an IoT, scenario has been presented yet.
Certainly, several aspects someway related to the scope have been considered in
the IoT or, in limited cases, in the SIoT scenario. As an example, in [
All the approaches mentioned above are extremely interesting; furthermore, several other related approaches proposed in the past to evaluate the relevance or the impact of a node in the Internet of Things could be mentioned. However, none of them has been conceived to operate in a complex scenario consisting of multiple IoTs, which interact with each other through smart and social objects that simultaneously belong to more of them and act as \bridges".
In this paper, we aim at providing a contribution in this setting by proposing an approach to computing the scope of a thing in a MIoT. Speci cally, after having provided an overview on the MIoT paradigm (Section 2), we present a de nition of scope along with a formalization allowing the computation of the corresponding values (Section 3). Then, we illustrate an application of scope in the context of smart cities, where it can play a key role (Section 4). Thereafter, we present some tests conceived to understand its main features (Section 5). Finally, we draw our conclusions and have a look at some possible future developments of our ideas (Section 6). 2
In this section, we provide an overview of the MIoT paradigm, described in detail
in [
Let oj be an object of M. We assume that, if oj belongs to Ik, it has an instance jk , representing it in Ik. The instance jk consists of a virtual view (or, better, a virtual agent) representing oj in Ik. For example, it provides all the other instances of Ik, and the users who interact with this IoT, with all the necessary information about oj . Information stored in jk is represented according to the format and the conventions adopted in Ik.
A MIoT M can be also represented by means of a graph-based notation. In particular, a graph Gk = hNk; Aki can be associated with an IoT Ik of M. In this case: { Nk is the set of the nodes of Gk; there is a node njk for each instance jk 2 Ik, and vice versa. Since there is a biunivocal correspondence between a node and an instance, in the following we shall use these two terms interchangeably. { Ak is the set of the arcs of Gk; there is an arc ajqk = (njk ; nqk ) if there exists a link from njk to nqk . A label can be associated with ajqk ; it stores the set tranSetjqk of the transactions performed from jk to qk in the past.
M = hN; Ai { N = Sm
k=1 Nk; { A = AI [ AC , where AI = Sm
k=1 Ak and AC = f(njk ; njq )jnjk 2 Nk; njq 2 Nq; k 6= qg.
AI is the set of the inner arcs (hereafter, i-arcs) of M; they link instances (of di erent objects) belonging to the same IoT. AC is the set of the cross arcs (hereafter, c-arcs) of M; they link instances of the same object belonging to di erent IoTs. A node connected to at least one c-arc is called c-node; otherwise, it is called i-node.
In M, an object oj has associated a set M Dj of metadata. Our metadata
model refers to the one of the IPSO (Internet Protocol for Smart Objects)
Alliance [
Now, we can de ne the set tranSetjk of the transactions activated by jk in Ik. Speci cally, let 1k ; 2k ; ; wk be all the instances belonging to Ik. Then: tranSetjk =
[ q=1::w;q6=j tranSetjqk
In other words, tranSetjk is given by the union of the sets of the transactions from jk to all the other instances of Ik. 3
In this section, we present the de nition of the scope of an instance jk in an IoT Ik and of the scope of an object oj in a MIoT. For this purpose, we must introduce some preliminary concepts.
The rst of them is the one of neighborhood of an instance jk in Ik. Specifically, the neighborhood nbh( jk ) of jk is de ned as:
nbhjk = out nbhjk [ in nbhjk in nbhjk = fnqk j(nqk ; njk ) 2 AI ; jtranSetqjk j > 0g
In other words, nbhjk comprises those instances directly connected to jk through an incoming or an outgoing arc, which shared at least one transaction with jk . In the following of this paper, we are interested only to out nbhjk ; as a consequence, when we will use the term \neighborhood", we will implicitly mean \out neighborhood".
Now, we need to introduce the neighborhood of level t of an instance jk in its IoT Ik. It is an extension of the concept of out nbhjk and is de ned as follows: where: and out nbhjk = fnqk j(njk ; nqk ) 2 AI ; jtranSetjqk j > 0g
out nbhtjk = out nbhjk f rk j rk 2 out nbhqk ; qk 2 out nbhtjk 1; rk 62 out nbhjwk ; 0 if t = 0 w < tg if t > 0
The concept of out nbhtjk will be extremely important later. In the meantime, we introduce a new concept, namely the one of minimum path jqk from an instance jk to an instance qk 2 nbhtjk . jqk is de ned as the succession f 0k ; 1k ; : : : ; tk g of instances such that 0k = jk , tk = qk , wk 2 out nbh(w 1)k for 1 w t.
After this, we must introduce the de nition of the Trust Degree T Dqjk of an instance qk in the instance jk in Ik. It can be de ned as the fraction of the transactions sent by jk to qk that have been requested by qk or that qk did not request but it has considered so interesting to repost or to elaborate them3. In order to formalize T Dqjk , we must introduce: { the set repostedqk of the transactions received by qk of Ik and reposted by it; { the set elaboratedqk of the transactions received by qk whose contents it elaborated for its purposes; { the set requestedqk of the transactions explicitly requested by qk .
If qk belongs to out nbhjk , T Dqjk can be expressed as:
T Dqjk = jtranSetjqk \ (requestedqk [ repostedqk [ elaboratedqk )j jtranSetjqk j
Starting from this de nition and from the concept of out nbhtjk , we can proceed with the transitive closure of T Dqjk in such a way as to extend it to the case in which qk is indirectly connected to jk . In particular, the general de nition of T Dqjk is as follows:
8 jtranSetjqk \(requestedqk [repostedqk [elaboratedqk )j if qk 2 out nbhjk T Dqjk = <>>> Qtw=1 T D((w 1)w)kjtranSetjqk j > : if qk 2 out nbhtjk ; t > 0; jqk = f 0k ; 1k ; ; tk g
The next step regards the de nition of the concept of Impact Degree IDjk of an instance jk in Ik. It is de ned as the average of the Trust Degrees that all the instances belonging to out nbhjk have in jk . It can be formalized as follows: IDjk =
P qk 2out nbhjk T Dqjk
jout nbhjk j
We are now able to de ne the Scope Sctjk of level t of an instance jk of Ik. Speci cally, Sctjk is de ned as the weighted sum of the Impact Degrees of the 3 Clearly, it might happen that an unrequested transaction of tranSetjqk is not considered interesting by qk . In this case, qk neither posts nor elaborates it. instances belonging to out nbhtjk , where the weights are the trust values that these instances have in jk . This sum is then averaged by the number of the instances belonging to nbhtjk . Formally speaking:
Sctjk =
P qk 2out nbhtjk T Dqjk IDqk
jout nbhtjk j
Now, we can de ne the Scope Sctj of level t of an object oj in the MIoT. It is obtained by averaging the Scope of level t of its instances in the corresponding IoTs. Speci cally, let Instj = f j1 ; j2 ; ; jl g be the instances of oj in the IoTs of the MIoT. Then:
Sctj =
P jk 2Instj Sctjk
jInstj j
From the de nitions of T Dqjk , IDqjk , Sctjk and Sctj it emerges that each of these parameters belongs to the real interval [0; 1]. 4
In a scenario characterized by the pervasive di usion of increasingly intelligent and social objects, our approach can have a large variety of applications. Two very interesting ones regard smart cities and shopping centers. Due to space limitations, we describe in detail only the rst one.
Consider some public areas (such as parks, squares, shopping centers, etc.) in a (smart) city, and assume that a set of people actively visits these areas. Each area is equipped with several smart objects for monitoring weather, air quality, tra c conditions, level of noise, etc., along with several actuators, such as smart lamps or information hubs provided as online services. Each person may be provided with several smart devices, such as smartwatches, smartphones, other wearable devices, and so forth. Persons and places can interact with each other through the corresponding smart objects.
Such a scenario can be modeled through a MIoT M consisting of a set fI1; I2; ; Img of IoTs, each representing a public area. The set of the objects of M comprises the smart objects installed in the public areas and the set of personal devices of people visiting them. If an object oj of the MIoT is active in the kth public area, it has an instance jk in the IoT Ik. Clearly, when a person with a smart object oj moves around di erent public areas corresponding to di erent IoTs, oj will have di erent instances, one for each IoT.
Each visitor of an area is generally interested to a certain kind of activity; for instance, she could be a tness runner. The nal goal of the MIoT is supporting people to get the best experience from their activities. In this setting, scope can play a role in reaching this objective. In the following, we report some possible usage scenarios.
Assume that a person wants to go out for a run. First, she needs to choose the best area for the run, based on weather conditions, tra c and other parameters that she considers relevant. To carry out her choices, she can check data provided by the sensors of each public area of her interest, the information hubs or other trusted runners. The choice of the data sources to consult is usually related to the corresponding trustworthiness and the easiness of getting desired information from them. These two properties are clearly strictly correlated to the scope of the source; indeed, this scope can be seen as a \summary" of these two parameters and some other related ones, such as accuracy, reputation, impact, etc. Once the person has performed her choice, she can send information to the MIoT in such a way as to serve, in her turn, as information provider for the community.
A similar activity ow may happen in several other circumstances, in which there is a decision to made, e.g., when a user must choose the best shopping center where she can buy a given object, the best cinema where she can see a movie, etc.
In all these cases, data regarding user choices can be coupled with those registered during the activities she performed therein (e.g., data coming from personal smart wears) in such a way as to con rm the correctness of the choice or, on the contrary, to alert the other users of the evaluation errors. For instance, imagine a scenario in which a person veri es that the weather was actually too cold for the clothes she had selected; interestingly, this information could be automatically detected and sent by the sensors in her smart wears. In this case, the scope of the smart wears is useful to understand how extended and how strong is its capability of in uencing the decision of the other users. In other words, the scope of an object oj in this scenario determines how many users are impacted by the data sent by it and how much strong this impact is.
It is worth pointing out the relevance of scope in this context. As a matter of fact, knowing the objects with the highest impact in the MIoT allows the improvement of the e ciency and the e ectiveness of the information disseminated through the network. At a higher abstraction level, some smart objects of the MIoT could assume the role of reliable information hubs for the whole MIoT if their scope is particularly strong and large over time. 5
In order to perform our experiments, since real MIoTs with the dimension and the variety handled by our model do not exist yet, we constructed a MIoT simulator. This tool starts from real data and returns simulated MIoTs with certain characteristics speci ed by the user.
The MIoTs created by our simulator follow the paradigm described in Section 2. Our simulator is also provided with a suitable interface allowing a user to \personalize" the MIoT to build by specifying the desired values for several parameters, such as the number of nodes, the maximum number of instances of an object, and so forth.
To make \concrete" and \plausible" the created MIoTs, our simulator leverages a real dataset. It regards the taxi routes in the city of Porto from July 1st 2013 to June 30th 2014. It can be found at the address http://www.geolink.pt/ ecmlpkdd2015-challenge/dataset.html. Each route contains several Points of Interests corresponding to the GPS coordinates of the vehicle.
We partitioned the city of Porto in six areas and associated a real IoT with each area. Our simulator associates an object with a given route recorded in the dataset and an object instance with each partition of a route belonging to an area. It creates a MIoT node for each instance and a c-arc for each pair of instances belonging to the same route. Furthermore, it creates an i-arc between two nodes of the same IoT if the length of the time interval between the corresponding routes is less than a certain threshold tht. The value of tht can be speci ed through the constructor interface. Clearly, the higher tht the more connected the constructed MIoT.
The MIoT for this experiment, which we constructed through our simulator, consisted of 1256 nodes. This number of nodes is much higher than the ones presently characterizing real MIoTs. However, we preferred to construct such a large MIoT because we think that, with the enormous development of the IoT, in the future there could exist MIoTs consisting of thousands of nodes. On the other side, we did not adopt larger MIoTs because they required excessive computation times without providing more knowledge on scope than the one we could have acquired with a MIoT of about 1000 nodes. The six IoTs of the MIoT had 128, 362, 224, 280, 98 and 164 nodes, respectively. The constructed MIoT is returned in a format that can be directly processed by the cypher-shell of Neo4J. The interested reader can nd the MIoT adopted in the experiments described here at the address http://daisy.dii.univpm.it/miot/datasets/scope.
We carried out all the tests presented in this section on a server equipped with an Intel I7 Quad Core 7700 HQ processor and 16 GB of RAM with the Ubuntu 16.04 operating system. To implement our approach we adopted: { Python, as programming language; { Neo4J (Version 3.4.5), as underlying DBMS.
In these experiments, we aimed at investigating the trend of the scope against the neighborhood level t. In particular, for each instance jk of the MIoT, we computed Sctjk when t increases from 1 to the diameter of Ik. After this, we grouped the instances of our MIoT in several ways (one for each test), based on some speci c rationales, and we computed the variation of the average values of the scope for each group.
As a rst task of this activity, we computed the variation of the average values of the scope for each IoT of the MIoT. This is equivalent to say that the instances of the MIoT were grouped in the corresponding IoTs (one group for each IoT). The obtained results are reported in Figure 1. From the analysis of this gure, we can observe that, in each IoT, the scope decreases quite quickly. Indeed, it is extremely high when t = 1 in all the IoTs. When t = 2, the scope is high for the largest IoTs, whereas it has an intermediate value for the other ones. In any case, the scope becomes very low when t is greater than or equal to 4 for small networks and when t is greater than or equal to 5 for the large ones.
As a second task, we computed the variation of the average values of the scope for the whole MIoT. This is equivalent to say that we had a unique group comprising all the instances of the MIoT. The obtained results are reported in Figure 2. From the analysis of this gure, we can conclude that the scope presents a trend similar to the one of the largest IoTs of Figure 1. In particular, it is very high for t = 1; it is high for t = 2; it has an intermediate value for t = 3, whereas it is very low for t > 5.
As a nal task, we subdivided all the instances of the MIoT in two groups containing i-nodes and c-nodes, respectively. Then, we computed the variation of the average values of the scope for the two groups. The nal goal of this task was to verify if i-nodes and c-nodes showed di erent behaviors as far as their scope is concerned. The obtained results are reported in Figure 3. From the analysis of this gure, we can observe that the scope decreases for both i-nodes and cnodes. However, the trends are di erent. Indeed, the decrease is much smother for i-nodes than for c-nodes. In particular, for c-nodes, the decrease is very steep because the scope is less than 0.2 already for t = 3. This can be explained by considering that, analogously to what was made in all the past approaches, our de nition of neighborhood (which plays a key role in our de nition of scope) considers as neighbors of a node only other nodes of the same IoT. In other words, it takes only i-arcs into account. Actually, we believe (and the results of Figure 3 represent a con rmation) that it is worthwhile to investigate the role of c-arcs in the computation of the neighborhood of a node and we plan to make this investigation in the future.
As for the investigation of the values of the scope for objects, we observe that they are obtained by averaging the values of the scope of the corresponding instances. As a consequence, it does not make sense to perform the rst and the nal tasks of the previous activity. The only task that makes sense is the second one; in this case, the variation of the average values of the scope of objects for the whole MIoT is reported in Figure 4. As we could have expected, this trend is very similar (or, better almost identical) to the one of Figure 2. 6
In this paper, we have seen that social internetworking and the Internet of Things are becoming more and more contiguous and are giving rise to several social and/or multiple IoTs paradigms. In this new scenario, we have introduced the concept of scope of a thing in a MIoT, along with a formalization allowing the computation of the corresponding values. Then we have illustrated one possible application along with some experiments aiming at evaluating the variation of scope against the value of the neighborhood level.
This paper should not be considered as an ending point. Instead, it could be the starting point of many researches in this eld. Indeed, there are several future related investigations that could be made in this context. First, we would like to analyze the role of possible constraints on network nodes or arcs in the de nition of scope. Then, we plan to study the role of scope in the detection of anomalies and, even more, for understanding the extension and the importance of the damage caused by them. Finally, we would like to analyze the exploitation of scope in predictive maintenance, that is currently one of the most important research issues in the manufacturing industry.