Stream Data Access Control Model: The Need for Data Semantics

Theppatorn Rhujittawiwat

, Ahmed Saaudi

and Csilla Farkas

University of South Carolina, Columbia, South Carolina, U.S.A.

University of Al-Muthanna, Samawh, Iraq

Keywords:

Stream Data, Ontologies, Dynamic Access Control, Semantics-based Authorization.

Abstract:

In this paper, we propose a semantics-based authorization model for stream data. We demonstrate that current

authorization models are insufﬁcient to provide dynamic access control for emerging technologies, such as

the Internet of Things. We propose an authorization model using ontologies and rules to express security

requirements for stream data. Our model supports secure interoperation and it is independent from the data

syntax. We propose security object patterns to express access control needs. These patterns are associated

with ontological concepts. Data instances inherit the security protection needs via stream pattern mapping.

Security object patterns can incorporate domain, context, and temporal security restrictions. We show that our

model is resistant to attacks that aim to bypass security restrictions by reconstructing stream data.

1 INTRODUCTION

Stream data management is an important require-

ment for emerging computing environments, such

as the Internet of Things (IoT). Initiatives, such

as IoT-Lite (Bermudez-Edo et al., 2015)(Bermudez-

Edo et al., 2016) and Stream Annotation Ontology

(SOA) (Kolozali et al., 2014), aim to support intel-

ligent interoperation. Along with the development of

IoT applications, concerns about cybersecurity threats

emerged. In this paper, we address the speciﬁc cyber-

security needs of stream data in the IoT environment.

We argue that current stream data security models do

not support dynamic security, thus, insufﬁcient to sup-

port IoT applications.

During the last decade, several researchers ad-

dressed the need to provide data security for stream

data. Most of these models address conﬁden-

tiality protection for stream data. The proposed

solutions range from encryption-based end-to-end

communication security (Penrig et al., 2000)(Puthal

et al., 2015)(Puthal et al., 2017)(Veltri et al.,

2013)(Zhu et al., 2006) to proposing generic secu-

rity models (Carminati et al., 2015)(Manogaran et al.,

2017)(Mengke et al., 2016)(Nehme et al., 2008).

Communication security for stream data is widely

provided by end-to-end encryption during transmis-

sion. To further improve the performance of this

method, many researchers introduced efﬁcient end-to-

end encryption techniques such as the dynamic key

length based approach (Puthal et al., 2015) and ap-

proaches for grouping and sharing keys (Penrig et al.,

2000)(Puthal et al., 2017)(Veltri et al., 2013)(Zhu

et al., 2006). However, end-to-end encryption tech-

niques provide conﬁdentiality over the entire data

stream and are unable to provide ﬁne granularity au-

thorization. These techniques are cumbersome to en-

force advanced security requirements, such as role-

based access control, context-aware security, and dy-

namic security policies.

Several researchers proposed access control pol-

icy models for stream data (Cao et al., 2009)(Carmi-

nati et al., 2010)(Carminati et al., 2007a)(Carminati

et al., 2007b)(Thoma et al., 2019). These approaches

adopt access control concepts of traditional database

management systems (DBMSs) to stream data. How-

ever, these efforts do not sufﬁciently address the

speciﬁc needs of data stream management systems

(DSMSs). For example, a DSMS can access only

incoming data while a DBMS can access the com-

plete data. The operator scheduling in a DSMS

signiﬁcantly impacts system performance. Operator

scheduling strategies have been proposed by Carmiati

et al. (Carminati et al., 2010) and by Thoma et

al. (Thoma et al., 2019).

An enforcement of security requirements over the

stream data instance has been proposed by Nehme et

al. (Nehme et al., 2008). The security requirements

are assigned to each attribute by their positions in the

stream tuple. Their technique, called security punc-

tuation, provides efﬁcient access control for stream

data. However, since the security punctuation is based

on the syntax of the stream data, this approach pro-

vides limited interoperability. Moreover, modiﬁca-

Rhujittawiwat, T., Saaudi, A. and Farkas, C.

Stream Data Access Control Model: The Need for Data Semantics.

DOI: 10.5220/0011528600003335

In Proceedings of the 14th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2022) - Volume 3: KMIS, pages 19-30

ISBN: 978-989-758-614-9; ISSN: 2184-3228

tion of the stream syntax may allow the attacker to

bypass the access control restrictions.

Malicious corruption of the integrity of stream

data was addressed by Guo et al. (Guo et al., 2007).

The authors proposed a watermarking method to de-

tect malicious modiﬁcation of the stream data. The

authors introduced the concept of data stream com-

pleteness. While their methods are able to detect

modiﬁcation of the data stream, they do not provide

a model to express protection requirements.

In this paper, we present a security model to

express security needs for stream data. Our work

is based on the intersection of stream data se-

mantics modeling and semantics-based authorization

models. Data semantics is widely used to sup-

port stream data integration and data analytics (Le-

Phuoc et al., 2011)(Whitehouse et al., 2006). Our

solution was also inﬂuenced by work to support

semantics-based authorization models (Sabelfeld and

Myers, 2003)(Choi et al., 2014) and context-based

decision support (Celdr

an et al., 2014)(Chen et al.,

2003)(Montanari et al., 2005).

The main contributions of our work are to 1) ex-

press security requirements via stream data semantics

and 2) support dynamic and ﬂexible policy manage-

ment. Our focus is on the concept of protected object

patterns. These patterns represent (partial) stream tu-

ples and their combinations. Values for the tuple at-

tributes may be variables or constants, thus allowing

for the expression of security needs over varying lev-

els of granularity. We model the mapping from the

protection objects to the data instances via the cor-

responding ontologies. A protection object may be

applicable to multiple ontology concepts and a con-

cept may have multiple protection objects mapped to

it. Stream data instances inherit the security require-

ments from the security object patterns. We show that

our security label assignment provides strong conﬁ-

dentiality protection. Syntactically different but se-

mantically same data items will have the same secu-

rity protection assignment. In this paper, we present

our functional model. We also give proof of formal

properties of our security label assignment.

The organization of the paper is as follows: Sec-

tion 3.2 discusses the motivating situations and prob-

lems which drive the direction of the solution. Section

4 deﬁnes the stream data model and security assign-

ment. Section 5.1 describes the concept of stream data

ontology. Section 5.2 illustrates logical reasoning and

how to incorporate it into security policy. Section 6

presents an example of security policy implementa-

tion using ontology. We conclude in Section 7.

2 RELATED WORK

In this section, we give an overview of current re-

search on stream data security, dynamic access con-

trol, and the limitations of these works.

2.1 Stream Data Security

Current work on stream data security focuses on con-

ﬁdentiality (Carminati et al., 2015)(Manogaran et al.,

2017)(Mengke et al., 2016). The ﬁrst priority of

stream data design is performance. To provide se-

curity capabilities with the least effect on perfor-

mance, the major methods of providing conﬁden-

tiality are based on end-to-end cryptography (Pen-

rig et al., 2000)(Puthal et al., 2015)(Puthal et al.,

2017)(Veltri et al., 2013)(Zhu et al., 2006). However,

end-to-end encryption techniques provide conﬁden-

tiality over only the whole package. They are unable

to provide ﬁne granularity authorization, and cannot

enforce advanced security requirements, such as role-

based, lattice-based access control, or dynamic se-

curity policies. Therefore, such requirements must

be enforced by the receiving applications. Many re-

searchers introduce syntax-based security techniques

for streaming data. The security requirements are

assigned to each attribute by their positions in the

stream tuple. Nehme et al. (Nehme et al., 2008) pro-

posed a technique called security punctuation to pro-

vide efﬁcient access control for stream data. A secu-

rity punctuation technique focuses on providing pri-

vacy control for the data source side. The current se-

curity policy is sent from the data source and speciﬁes

the security labels of attributes in the data tuple. This

approach allows the data source side to decide the se-

curity needs which are suitable for privacy sensitive

data domains such as medical information.

The growing ﬁeld of IoT has led to increasing

number of stream data applications. Many of these

applications also collect sensitive user data. To pro-

tect conﬁdentiality and privacy of users, researchers

propose access control for IoT stream data in differ-

ent context. For example, Shafagh et al. (Shafagh

et al., 2020) proposed a decentralized access control

based on blockchain technology. Their approach pro-

vides an extra layer of anonymity for users but has

a limited security granularity. Nambiar et al. (Nam-

biar et al., 2020) extended Apache Storm to protect

privacy of the data in distributed systems. Their ap-

proach supports a variety of access control models,

such as Discretionary Access Control (DAC), Manda-

tory Access Control (MAC), and Role-Based Access

Control (RBAC).

Semantic approaches have been used in stream

KMIS 2022 - 14th International Conference on Knowledge Management and Information Systems

data since the ﬁeld was established. However, their

main focus is on how to interpret data and use it

for data analytics (Le-Phuoc et al., 2011)(Whitehouse

et al., 2006)(Sejdiu et al., 2021). The security com-

munity has demonstrated the beneﬁt of semantics-

based security, such as Sabelfeld et al. (Sabelfeld

and Myers, 2003)(Choi et al., 2014)(Celdr

an et al.,

2014)(Chen et al., 2003)(Montanari et al., 2005).

However, the potential of these approaches for sup-

porting stream data security and privacy has not been

fully investigated.

Moreover, stream data requires dynamic security

policies because it is time sensitive. The same at-

tribute in stream data may require different security

levels at different times such as healthcare informa-

tion in emergencies. Lu et al. (Lu et al., 2012)

proposed a framework to provide privacy on mo-

bile health monitoring devices. This approach uses

mathematical models on monitored information to

detect emergency situations and to release informa-

tion to qualiﬁed helpers. Other mathematical models

to express emergency situations were also proposed

by many researchers (Jeong et al., 2014)(Lee et al.,

2013)(Yu et al., 2017).

2.2 Dynamic Security Policy

Privacy is the major security concern related to com-

mercial IoT devices. Recent research shows that cur-

rent IoT devices focus on having secure connections

between devices and the server (Alharbi and Aspinall,

2018)(Janes et al., 2020)(Ren et al., 2019). However,

they lack a good access control policy to protect sen-

sitive information. The information can be exposed to

the third-party that operates the service such as Ama-

zon Web Services (AWS) (Ren et al., 2019). End

users, such as smart home users, are often left with

little or no control over device security. Moreover,

security policies may be dependent on the context.

Semantics-based approaches can be used to express

the context and situations, and adjust the security pol-

icy accordingly.

Carmiati at al. (Carminati et al., 2015) propose

an approach to provide privacy while working around

the current static model, such as using statistic mod-

els to detect anomalies in data and granting access to

all data if they are requested during the presence of an

anomaly. Semantics-based security policies can pro-

vide needed dynamic capabilities to protect privacy

without hindering emergency management. An ar-

chitecture that incorporates semantics-based security

policy into a stream data model has the potential to

provide a dynamic security policy that provides func-

tionalities in emergencies without sacriﬁcing privacy.

2.3 Vulnerabilities of Stream Access

Control Enforcement via Syntactic

Policy Enforcement

Nehme et al. (Nehme et al., 2013)(Nehme et al., 2008)

proposed a security policy for stream data called se-

curity punctuation (sp). The security punctuations

are inserted by the source of the stream to indicate

the protection needs of the stream data. A security

punctuation contains the following: punctuation type

(pt), data description part (ddp), security restriction

part(srp), sign, timestamp (ts), and enforcement (et).

The security punctuation technique focuses on pro-

viding privacy control for the data source side. The

current security policy is sent from the data source

and speciﬁes the security labels of attributes in the

data tuple. This approach allows the data source side

to decide the security needs which are suitable for pri-

vacy sensitive data domains, such as medical informa-

tion.

Example 1: Security Punctuation Policy Examples.

1 . D at a S e c u r i t y P u n c t u a t i o n

< dsp|S

|+|1 : 00 : 00PM|I >

Only q u e r i e s r e g i s t e r e d by r o l e R

can q u e r y

t h e s t r e a m S

< dsp|A

, A

|+|1 : 00 : 00PM|D >

Only q u e r i e s r e g i s t e r e d by r o l e R

can q u e r y

t h e a t t r i b u t e s A

an d A

a f t e r 1PM.

2 . Qu ery S e c u r i t y P u n c t u a t i o n

< qsp|null|R

|+|1 : 00 : 00PM|I >

Qu ery a c q u i r e s a r o l e R

< qsp|S

|+|1 : 00 : 00PM|D >

Qu ery a c q u i r e s a r o l e R

a f t e r 1PM a nd t h e r o l e

i s p e r m i t t e d t o o n l y a c c e s s s t r e a m S

The security punctuation is subject to attacks that

can destroy or delay the sp. This is a common vul-

nerability of syntax-based security policies. Mali-

cious attackers can potentially disclose unauthorized

data by changing the position of the target tuple or

attribute, thus making the system interpret the secu-

rity level incorrectly. For example, due to the na-

ture of DSMS, the data can be accessed without au-

thorization if the security policy is not updated with

the correct time. The new sp can apply to only tu-

ples which are still in the sliding window and tuples

which arrive after sp. However, it cannot apply to tu-

ples which are ﬁnished querying and not in the sliding

window anymore. In the case that sp is destroyed dur-

ing transmission, the policy will not be updated. This

gives opportunities for attackers to access the unau-

thorized data with the outdated policy. In the case

that sp is delayed during transmission, attackers can

access the unauthorized data with the outdated policy

until sp arrives. Moreover, if the data items are shuf-

ﬂed, the security punctuation that refers to the data

Stream Data Access Control Model: The Need for Data Semantics

Figure 1: Architecture.

position in the tuple, may indicate the incorrect secu-

rity requirements. Our work, that describes security

for stream tuples based on data semantics, is indepen-

dent of the actual stream syntax and can avoid these

attacks. Moreover, our approach can be used with the

security punctuation-based implementation to verify

that the security has not been compromised.

3 ACCESS CONTROL

ARCHITECTURE AND

MOTIVATING EXAMPLE

The main focus of this work is to represent syntax-

independent access control requirement for stream

data. In this section, we present the system architec-

ture for our stream data access control and a motivat-

ing example demonstrating the need for a semantics-

based security model.

3.1 Architecture

We propose the use of ontologies to model domain

and context semantics. Security requirements are ex-

pressed over the ontologies and dynamically updated

according to context changes. Figure 1 shows the

overview of our system architecture. The policy part

of architecture contains domain and context ontolo-

gies, the reasoning engine, and the list of security ob-

jects. The policies will be applied on data instances of

the stream. The reasoning engine will derive security

labels from ontologies and assign them to security ob-

jects. The instances matching the security object de-

scriptions will be assigned by security labels of those

security objects.

3.2 Motivating Example

In this section, we present motivating examples to

demonstrate the shortcomings of current stream data

authorization models.

Figure 2: Oxygen saturation monitoring results on a patient

in an emergency ward.

In stream data, only a part of the data can be pro-

cessed at any given time via continuous queries. The

data will be processed with the same queries over

and over again. A different order of the same stream

tuples may result in different computational results.

Similarly, missing tuples will impact the results of

the continuous queries. Incorrect ordering can be de-

tected in case the timestamps are correct. However,

malicious users may modify the timestamp to achieve

a desired evaluation schedule. Moreover, the com-

pleteness of the stream may be difﬁcult to detect if

there is no predetermined schedule for the arrival of

the stream tuples. Next, we present a hypothetical

scenario that may occur in the presence of malicious

users.

Incorrect Oxygen Saturation Measurement.

Health monitoring sensors broadcast vital signs,

such as body temperature, pulse rate, respiration

rate, blood pressure, and pulse oximetry. These

vital signs are frequently observed by medical staffs

during hospital admission. The frequency can vary

depending on the condition of patients. For example,

pulse oximetry measures the oxygen saturation in the

blood. The normal value is between 95-100%. An

oxygen saturation value below 90% is referred to as

hypoxemia. Patients generally show signs of mental

impairment at oxygen saturation below 85% and

completely lose consciousness at below 75%(Powell

et al., 1996). The presence of hypoxemia is generally

caused by underlying conditions of the heart or lungs,

which may require emergency care. The lack of

oxygen from hypoxemia can lead to a wide range of

complications from minor headaches to respiratory

failure, which can lead to death or permanent brain

damage in a matter of minutes without an immediate

response. The oxygen saturation level is generally

observed every 4 hours in admitted patient care but

constantly observed in critically ill patient cases.

Figure 2 shows an example of oxygen saturation

monitoring results. In this example, as soon as hy-

poxemia was detected, treatment was initiated. The

timely treatment prevented long-term complications

for the patient.

KMIS 2022 - 14th International Conference on Knowledge Management and Information Systems

Figure 3: Incorrect order of monitoring results.

Figure 3 shows an example of the same situation

as Figure 2 with an incorrect ordering. The medical

staff missed the early low oxygen saturation sign at

9:10 AM and didn’t initiate the emergency procedure.

To avoid penalty, they switched the timestamp of the

measurements, claiming that the readings were incon-

clusive. The patient did not receive appropriate treat-

ment in time which lead to long-term complications.

4 STREAM DATA MODEL

In addition to traditional data security requirements,

stream data requires security capabilities to provide

semantics-based dynamic security policy. Incomplete

data and incorrect order may affect the integrity and

security of the stream data. They also cannot be ad-

dressed with relational database security approaches.

A semantics-based dynamic security policy is needed

to provide fast and situation-responsive availability.

It can also counter attribute shufﬂe attacks which are

unique to the stream data model due to the lack of

metadata in tuples.

In this section, we present our framework to

address the above requirements.

Deﬁnition 1. Data Schema

A data schema D is a set of attribute names denoted by

D(A

, ..., A

), where D is the name of the data schema

and A

|i ∈ {1, ..., n} is an attribute name.

Deﬁnition 2. Data Instance on Data Schema D

A data instance d on data schema D(A

, ..., A

) is de-

noted as d =< a

, ..., a

>, where a

|i ∈ {1, ..., n} is

the value of attribute A

such that a

is in the domain

of attribute A

, i.e., a

∈ Dom(A

We also denote data instance d as d =< A

, ..., A

= a

> to explicitly identify the attribute val-

ues.

Deﬁnition 3. Stream Tuple

A stream tuple t from source l is denoted as t =

(l, d, ts), where ts is a timestamp representing the time

when t was generated.

Example 2: A stream tuple t

from source sensor

1:00AM

= ( sensor

, < A

= 10, A

= 20, A

= 30 > , 1 : 0 0 : 0 0AM)

The above stream tuple t

was generated by source

sensor

at 1:00:00AM. It contains a data instance

d =< A

= 10, A

= 20, A

= 30 > which specify the

value of attribute A

is 10, the value of attribute A

20, and the value of attribute A

is 30.

Deﬁnition 4. Data Stream

A data stream S from source l during a time period

from ts

to ts

is denoted as S

=< t

, ..., t

>, where

= (l

, d

,ts

) such that l

= l and < ts

, ..., ts

> is a

sequence of timestamps during a time period from ts

to ts

We also denote a data stream S from source

l during time interval from ts

to ts

[ts

,ts

] =< t

, ..., t

> or S

[ts

,ts

] =<

(l, d

,ts

), ..., (l, d

,ts

) >.

Deﬁnition 5. Stream Bundle

Let ts denotes a timestamp and l

, ..., l

denote stream

data sources. A stream bundle at the time ts, denoted

= {(l

, d

,ts), ..., (l

, d

,ts)}, is the set of stream

tuples generated by the sources l

, ..., l

at time ts.

Note, at this point we do not address the issue of in-

compatible time stamps due to different levels of pre-

cision of the sources or due to the lack of synchroniza-

tion.

Deﬁnition 6. Atomic Protected Object Pattern

An atomic protected object pattern, also referred to

as an atomic protected object, is a triple consisting of

3 elements; a source description, a data description,

and a timestamp description. A protected object o is

denoted as

o = ((l, exp

), (d, exp

), (ts, exp

))

where each element is as follows. Let v denotes a

variable, c denotes a constant, A

an attribute name

in d, and op denotes the operation = or <. Note,

we require that all attributes of the expressions are

bounded, i.e., they must appear in the corresponding

element l, d or ts.

1. A source description (l, exp

) where

(a) l is a data source c, such that c is a valid source,

or v;

(b) exp

is a Boolean expression of the form

Component

AND ... AND Component

, where

Component

(i = 1, . . . , n) is T RUE or (v op c).

We also denote a source description (v, T RUE),

representing any source, as (∗).

Stream Data Access Control Model: The Need for Data Semantics

Figure 4: Examples of different kinds of protected objects;

1. Atomic Protected Object:A tuple, 2. Stream Bundle:

Tuples from multiple sources which arrive at the same time

period, 3. Data Stream: Multiple tuples which continuous

arrive from a single source

2. A data description (d, exp

) where

(a) d is a data instance on data schema D, denoted

as d =< A

= k

, ..., A

= k

>, where k

(i =

1, . . . , n) is c, such that c ∈ Dom(A

) or v;

(b) exp

is a Boolean expression of the form

Component

AND ... AND Component

, where

Component

(i = 1, . . . , n) is T RUE, (v

op c),

op v

), (A

op c), (A

op v

), or (A

op A

We also denote a data description (< A

, ..., A

= v

>, T RUE), representing any data

instance on data schema D, as (∗).

3. A timestamp description (ts, exp

) where

(a) ts is a timestamp, such that ts is c or v;

(b) exp

is a Boolean expression of the form

Component

AND ... AND Component

, where

Component

is T RUE or (v op c).

We also denote a timestamp description

(v, T RUE), representing any timestamp, as

(∗).

Deﬁnition 7. Simple Object Valuation

Let ν be a symbol mapping, such that;

1. ν maps a constant c to itself, i.e., ν : c −→ c.

2. ν preserves equality of variables, i.e., ν : v −→ c

and ν : v −→ c

then c

= c

3. ν maps special symbols to themselves, i.e.,

ν :=−→=, ν :<−→<.

4. ν maps an expression ν(x op y) = ν(x) op ν(y),

where x, y are any of the symbols above.

Deﬁnition 8. Stream Pattern Mapping

Let o = ((l, exp

), (d =< A

= k

, ..., A

= k

, exp

(ts, exp

)) be an atomic protection object, and ν a

symbol mapping. A stream pattern mapping N is de-

ﬁned as;

1. N(l, exp

) is deﬁned as

N(l, exp

) = (ν(l), ν(exp

)).

2. N(< A

= k

, ..., A

= k

>, exp

) is deﬁned as

N(< A

= k

, ..., A

= k

>, exp

)) = (< ν(A

) =

ν(k

), ..., ν(A

) = ν(k

) >, ν(exp

))

and N(A

) = N(k

) = ν(k

3. N(ts, exp

) is deﬁned as

N(ts, exp

) = (ν(ts), ν(exp

)).

Deﬁnition 9. Stream Pattern Mapping Satisﬁability

Let o be an atomic protected object. We say that the

stream tuple t satisﬁes o iff there is a pattern mapping

N from o to t denoted as N(o) = t, such that N(exp

)

= T RUE, N(exp

) = TRUE, and N(exp

) = TRUE.

Note, from now on, we only consider satisﬁed map-

ping when we use the phrase “pattern mapping”.

Example 3: Protected Object Mapping.

Giv en p r o t e c t e d o b j e c t o

an d t u p l e s t

, t

= ( ( sensor

,TRUE ) , ( < A

= v

, A

= 20 > ,v

< 20

AND 10 < v

) , ( v

,TRUE ) )

= ( sensor

, < A

= 15, A

= 20 > , 2 : 0 0 : 0 0AM)

= ( sensor

, < A

= 10, A

= 20 > , 2 : 0 5 : 0 0AM)

Mapping o

t o t u p l e t

1 . The v a l u a t i o n fro m s o u r c e d e s c r i p t i o n (l, exp

)

t o s o u r c e l

ν(sensor

) = sensor

ν(exp

) = TRUE

C o n d i t i o n s i n 1 . a r e s a t i s f i e d .

2 . The v a l u a t i o n fro m d a t a d e s c r i p t i o n (d, exp

) t o

d a t a i n s t a n c e d

ν(v

) = 15

ν(20) = 20

ν(exp

) = ν(v

)<20 AND 10<ν(v

) → 15<20 AND 10<15

ν(exp

) = TRUE

C o n d i t i o n s i n 2 . a r e s a t i s f i e d .

3 . The v a l u a t i o n fro m t u p l e d e s c r i p t i o n

(ts, exp

) t o t i m e s t a m p ts

ν(v

) = 2 : 0 0 : 0 0AM

ν(exp

) = TRUE

C o n d i t i o n s i n 3 . a r e s a t i s f i e d .

A l l v a l u a t i o n s a r e s a t i s f i e d . t

i s a member o f o

Mapping o

t o t u p l e t

1 . The v a l u a t i o n fro m s o u r c e d e s c r i p t i o n (l, exp

)

t o s o u r c e l

ν(sensor

) = sensor

ν(exp

) = TRUE

C o n d i t i o n s i n 1 . a r e s a t i s f i e d .

2 . The v a l u a t i o n fro m d a t a d e s c r i p t i o n (d, exp

) t o

d a t a i n s t a n c e d

ν(v

) = 10

ν(20) = 20

ν(exp

) = ν(v

)<20 AND 10<ν(v

) → 10<20 AND 10<10

ν(exp

) = FALSE

C o n d i t i o n i n 2 . i s n o t s a t i s f i e d .

3 . The v a l u a t i o n fro m t u p l e d e s c r i p t i o n

(ts, exp

) t o t i m e s t a m p ts

ν(v

) = 2 : 0 5 : 0 0AM

ν(exp

) = TRUE

C o n d i t i o n s i n 3 . a r e s a t i s f i e d .

V a l u a t i o n 2 . i s n o t s a t i s f i e d . t

i s n o t a member o f o

Deﬁnition 10. Object Dominance

Let o

and o

protected objects. We say that o

dom-

inates o

if for every tuple t, there is a valuation

ν(o

) = t, there must be a valuation ν(o

) such that

ν(o

) = t. We denote o

dominates o

as o

⊇

Object dominance is reﬂective, transitive, and asym-

metric, i.e., if o

⊇

and o

⊇

then o

⊇

5 SECURITY MODEL

We assign the default security labels for each entity

on ontologies. Those default security labels represent

KMIS 2022 - 14th International Conference on Knowledge Management and Information Systems

Figure 5: Ontology with Security Labels.

Figure 6: Stream Ontology.

security labels in the normal situation. We use a con-

text reasoner to decide what is the current situation

and assign an appropriate security label according to

that situation. In this section, we give a brief overview

of the ontologies and context reasoner.

5.1 Stream Data and Security Ontology

We expand the stream data ontology introduced by

Kolozali et al. (Kolozali et al., 2014), adding data se-

curity. Traditional access control models consider se-

curity to be a passive property of data quality. How-

ever, security is a complex entity associated with

many parts of the stream data, e.g., security capabili-

ties that devices can provide, security associated role

of users, and security property of data. We built an

ontology containing six main modules: Stream Data,

Device, Quality, User, Event, and Security.

Stream ontology is generalized so it can represent

the stream data environment. We extend the stream

ontology to represent the context of the stream data.

The transmitted data may have different meanings

in different contexts. We develop additional speciﬁc

context ontologies for the data domain of each indi-

vidual system. We further extend the stream ontology

with the context ontology.

We present an extensible context ontology for

modeling context in stream data environments. We

follow the context model from CONON (Wang et al.,

Figure 7: Context Ontology (Preuveneers et al., 2004) (Qin

et al., 2007) (Wang et al., 2004).

2004) and divide our context ontology into upper on-

tology and speciﬁc ontology. Our upper ontology

contains all basic contextual entities from CONON

with additional entities from CoDAMoS (Preuveneers

et al., 2004) and Smart Space (Qin et al., 2007). The

upper ontology describes 6 basic concepts: Com-

pEntity, Location, Person, Activity, Environment, and

Time. The speciﬁc ontology represents the details of

general concepts in each sub-domain.

5.2 Context Reasoning

To illustrate the logical reasoning mechanism, we

present a scenario in which the data generated from

sensors can have different levels of security poli-

cies depending on the current situation. For exam-

ple, when people are sleeping, their oxygen satura-

tion levels will naturally decrease because their body

needs less oxygen. Their bodies will reduce the res-

piration rate and heart rate during sleep. While the

oxygen saturation will also decrease during exercise,

this happens due to the body consuming more oxy-

gen. Despite having low oxygen saturation, the res-

piration rate and heart rate will increase during exer-

cise. An oxygen saturation value during rest that is

below 90% is considered a medical condition called

hypoxemia. However, an oxygen saturation value of

88% can be considered normal during intense exer-

cise. In this case, the oxygen saturation alone cannot

determine the condition of hypoxemia; the respiration

rate and heart rate must also be considered. We want

to provide a dynamic security policy to protect pa-

tient A’s privacy by limiting access during normal sit-

uations but allowing access during emergencies. By

using the context and ontology rules, we can derive

that the current oxygen saturation is normal or is indi-

cating an emergency situation. The security label as-

signments can be deﬁned as λ(o

) = TopSecret and

λ(o

) = Public where o

is an oxygen saturation dur-

ing a normal situation and o

is an oxygen saturation

during an emergency situation. The protected objects

Stream Data Access Control Model: The Need for Data Semantics

can be separated into 2 cases; rest and exercise and be

deﬁned as the following example;

Example 4: Protected Objects in Different Situations.

A p r o t e c t e d o b j e c t o

r e p r e s e n t i n g a nor m a l

s i t u a t i o n d u r i n g r e s t i n g

= ( ( s o u r c e = patient

,TRUE) , ( < oxygenSaturation = v

respirationRate = v

, heartRate = v

> , 90 < v

) ,

)

A p r o t e c t e d o b j e c t o

r e p r e s e n t i n g a nor m a l

s i t u a t i o n d u r i n g e x e r c i s i n g

= ( ( s o u r c e = patient

,TRUE) , ( < oxygenSaturation = v

respirationRate = v

, heartRate = v

> , 88 < v

AND 40 < v

AND

150 < v

) ,

)

A p r o t e c t e d o b j e c t o

r e p r e s e n t i n g an e m e rgenc y

s i t u a t i o n d u r i n g r e s t i n g

= ( ( s o u r c e = patient

,TRUE) , ( < oxygenSaturation = v

respirationRate = v

, heartRate = v

> , v

< 90 AND v

< 40 AND

< 150 ) ,

)

A p r o t e c t e d o b j e c t o

r e p r e s e n t i n g an e m e rgenc y

s i t u a t i o n d u r i n g e x e r c i s i n g

= ( ( s o u r c e = patient

,TRUE) , ( < oxygenSaturation = v

respirationRate = v

, heartRate = v

> , v

< 88 AND 40 < v

AND

150 < v

) ,

)

Deﬁnition 11. Ontology Mapping

Let O denote the ontology entity, and o =

((l, exp

), (d, exp

), (ts, exp

)) be an atomic protec-

tion object, and ν a symbol mapping. An ontology

mapping N is deﬁned as;

1. If O is a CompEntity concept entity, N(O) = o =

((l, exp

), (d, TRU E), (ts, T RUE) where (l, exp

)

is the source corresponding to compEntity O

with its associated location concept, and both

(d, T RUE), (ts, T RU E) can be any given data de-

scription and timestamp description.

2. If O is either person, activity, or en-

vironment concept entity, N(O) =

o = ((l, T RUE), (d, exp

), (ts, T RUE) where

(d, exp

) is the attributes corresponding to either

person, activity, or environment concept asso-

ciated with O, and both (l, T RU E), (ts, T RUE)

can be any given source description and data

description.

3. If O is a time concept entity, N(O) =

o = ((l, T RUE), (d, T RUE), (ts, exp

)) where

(ts, exp

) is the timestamp corresponding to

time concept O, and both (l, T RU E), (d, T RU E)

can be any given source description and data

description.

4. If O is an ontology entity associated

with more than one concept, N(O) =

o = ((l, exp

), (d, exp

), (ts, exp

)) where

(l, exp

), (d, exp

), (ts, exp

) are the source,

attributes, and timestamp corresponding to

ontology entity O.

Deﬁnition 12. Protection Object Security Label As-

signment

Let O

Dom

denote the domain ontology, and Λ the

mapping of security labels to the concepts in O

Dom

We denote Λ(O

Dom

) the assignment of security la-

bels to all the concepts in the ontology, and λ(ct

)

to denote the security label of the concept ct

. We

require that λ(ct

) ≥ λ(ct

) if ct

is a descendent

of ct

). The set of Core Security Objects (CSO)

is the set of all object patterns corresponding to

the ontological concepts. That is, for CSO =

(ct

, T RU E), (ct

, T RU E), Λ(CSO) =

LUB(λ(ct

), λ(ct

)).

Additional security restrictions due to context-

aware or temporal needs, may be expressed over

more speciﬁc protection objects. Let o

⊇

, then

Λ(o

) ≤ Λ(o

Note, that emergency scenarios may require mod-

iﬁcation of core security object classiﬁcations and, in

our current model, can only be handled by reassigning

labels to the ontological concepts.

Deﬁnition 13. Instance Label Assignment

Let t be a data stream tuple and N(o

), ..., N(o

)

be satisﬁable pattern mapping from protected objects

, ..., o

to t, where

λ(o

), ..., λ(o

) are the security labels of o

|i ∈

{1, ..., k}. We say that the security label of t

is LUB(λ(o

))|i ∈ {1, ..., k} where LUB(λ(o

)) is

the lowest upper bound of the security labels

λ(o

), ..., λ(o

Algorithm 1: Assign security label to stream tuple.

Require: Stream tuple t, and security labels of all

protected objects λ(o

), ..., λ(o

)

Ensure: Label assignment to a stream tuple

1: Initialization

2: Find all pattern mappings N

, ..., N

from O (all

protected objects) that are satisﬁed by t

3: Let {o

, ..., o

} ∈ O be the protected objects that

can be mapped by N

, ..., N

to t

4: Let λ(t) = LUB(λ(o

), ..., λ(o

))

Example 5: Security Label Assignment to Stream Tu-

ple.

S e c u r i t y L a b e l s : T o p S e c r e t ( TS ) > S e c r e t ( S ) > P u b l i c ( P )

= ( ( sensor

, TRUE) , < A

= 20, A

= 20 > , 2 : 0 0 : 0 0AM)

CASE 1 :

= ( ( sensor

,TRUE ) , ( < A

= v

, A

= 20 > ,v

< 50 ) ,

( v

,TRUE ) )

= ( ( sensor

,TRUE ) , ( < A

= v

, A

= 20 > ,v

< 10 ) ,

( v

,TRUE ) )

λ(o

) = Secret

λ(o

) = To pSecret

A s s i g n s e c u r i t y l a b e l t o t

KMIS 2022 - 14th International Conference on Knowledge Management and Information Systems

i s mapped t o o

L a b e l o f t

= λ(o

) = S e c r e t .

CASE 2 :

= ( ( sensor

,TRUE ) , ( < A

= v

, A

= 20 > ,v

< 50 ) ,

( v

,TRUE ) )

= ( ( sensor

,TRUE ) , ( < A

= v

, A

= 20 > ,v

< 30 ) ,

( v

,TRUE ) )

λ(o

) = Secret

λ(o

) = To pSecret

A s s i g n s e c u r i t y l a b e l t o t

i s mapped t o b o t h o

an d o

LUB(λ(o

), λ(o

)) = λ(o

)

L a b e l o f t

= λ(o

) = Top S e c r e t .

Next, we propose the concept of a complex pro-

tected object. This allows to express protection needs

over aggregates of stream tuples, while allowing max-

imum availability of the individual tuples.

Deﬁnition 14. Complex Protected Object

Let o

, ..., o

be the atomic protected objects. We say

that data stream S =< o

, ..., o

> or stream bundle

B = {o

, ..., o

} is a complex protected object O

iff

λ(O

) = λ(S) > LUB(λ(o

))|i ∈ {1, ..., n} or λ(O

) =

λ(B) > LU B(λ(o

))|i ∈ {1, ..., n}.

is minimal, that is for any o

= O

- o

then

λ(

) ≯ LUB(λ(o

))| j ∈ ({1, ..., n}−{i}). And there

are no two objects o

and o

such that o

⊇

Theorem 1. Given a stream tuple t and security la-

bels of all protected objects λ(o

), ..., λ(o

). Algo-

rithm 1 correctly assigns a security label to a stream

tuple t such that the assigned label l is the LUB of all

labels of security objects that can be mapped to t.

Proof. By contradiction;

Let λ(o

) be a security label assigned to a stream tu-

ple t. Assume that λ(o

) does not satisfy the security

requirement of t.

Case 1: o

cannot be mapped to t

1. From algorithm 1, o

must be a protected object

that can be mapped to t.

2. So t must be a member of o

. This contradicts the

supposition that t is not a member of o

Case 2: Security label λ(o

) is lower than the security

requirement of t

1. There must exist a o

such that ν(o

) > ν(o

) and

there is a pattern mapping from o

to t, ν(o

) = t.

2. But the algorithm 1, λ(t) is assigned by

LUB(λ(o

), ..., λ(o

)) where {o

, ..., o

}

be the protected objects that can be mapped to t.

3. So λ(o

) must be LU B(λ(o

), ..., λ(o

)) which is

the highest security labels of protected objects

mapped to t. This contradicts the supposition that

λ(o

) is lower than the security requirement of t

Case 3: Security label λ(o

) is higher than the secu-

rity requirements of t

1. From algorithm 1, λ(t) is assigned by

LUB(λ(o

), ..., λ(o

)) where {o

, ..., o

} be

the protected objects that can be mapped to t.

2. So λ(o

) must be LU B(λ(o

), ..., λ(o

)) which is

the highest security labels of protected objects

mapped to t. This contradicts the supposition that

λ(o

) is higher than the security requirement of t

6 IMPLEMENTATION

In our previous work (Rhujittawiwat et al., 2021),

we implemented our context-aware security policy

engine. We implemented our framework using

Prot

e (Stanford, 2022) and Semantic Web Rule

Language (SWRL) on an open-source home automa-

tion called Home Assistant (Home Assistant, 2021).

Figure 8 shows a simpliﬁed ontology containing

StreamTuple and ProtectedObject and ﬁgure 9 shows

a protected object with simple elements describing the

source, data instance, timestamp, and security label.

SWRL allows us to write reasoning rules to assign se-

curity labels where protected object o = ((source =

l, exp

), (d, exp

), (timestamp = ts, exp

)) with label

assignment λ(o) = Label in the form as following;

StreamTuple(?t)

∧ hasSource(?t, N(source = l, exp

))

∧ hasDataInstance(?t, N(d, exp

))

∧ hasTimestamp(?t, N(timestamp = ts, exp

))

→ hasLabel(?t, Label)

(1)

Where N(source = l, exp

), N(d, exp

), and

N(timestamp = ts, exp

) can be derived from;

(2)

ProtectedOb ject(? o)

∧ source(?o, ?source)

∧ dataInstance(?o, ?data)

∧ timestamp(? o, ?timestamp)

∧ StreamTuple(?t)

∧ hasSource(?t, ?source)

∧ hasDataInstance(?t, ?data)

∧ hasTimestamp(?t, ?timestamp)

We show the SWRL rule to assign a security label

in ﬁgure 10 and the result where the label is assigned

to a stream tuple in ﬁgure 11

We use XML to facilitate data exchange between

Home Assistant and Prot

e. This approach makes

it possible to extend data sharing for future applica-

tions. Figure 12 outlines how the data is extracted

Stream Data Access Control Model: The Need for Data Semantics

Figure 8: The Stream Ontology in the Prot

Figure 9: The Protected Object in the Prot

Figure 10: Label assignment with SWRL in the Prot

Figure 11: Reasoning result from label assignment.

from the Home Assistant database, converted to the

XML schema, and shared with Prot

e. Similarly,

the decision reached by Prot

e is converted to the

XML format, and incorporated into the Home Assis-

tant database.

Figure 12: Communication between Home Assistant and

Prot

e via XML (Rhujittawiwat et al., 2021)

To get data from Prot

e into Home Assistant,

we converted the CSV ﬁle that was exported from

Prot

e. We created bash scripts to convert data from

each side to XML and convert XML to respective data

formats for each sides. Our ongoing work addresses

the security label enforcement using security punctu-

ation.

7 CONCLUSION

In this paper, we proposed a semantics-based access

control model for stream data. Our solution elimi-

nates the dependency on stream syntax and provides

1) syntax independent expression of security require-

ments, 2) high assurance of security compliance over

the IoT network, and 3) dynamic and adaptable se-

curity policy representation. We proposed the con-

cept of a stream data protection object that allows a

user-friendly and inter-operable expression of the pro-

tected data. We developed techniques so stream data

instances can inherit security requirements from the

protection object. In our model, we support strong

security by enforcing the most restrictive security re-

quirement that an instance inherits. We also cou-

pled our security model with context-based security.

The security restrictions of the protection objects may

change based on the current context. Our work pro-

vides an approach to support data security for the

rapidly evolving IoT environment. We applied tech-

nologies that are widely used for IoT applications to

support data integration and intelligent analysis, thus

our model is compliant with the technologies already

used by the industry.

REFERENCES

Alharbi, R. and Aspinall, D. (2018). An iot analysis frame-

work: An investigation of iot smart cameras’ vulnera-

bilities. In Living in the Internet of Things: Cyberse-

curity of the IoT - 2018. IET.

KMIS 2022 - 14th International Conference on Knowledge Management and Information Systems

Bermudez-Edo, M., Elsaleh, T., Barnaghi, P., and Taylor, K.

(2015). Iot-lite ontology.

Bermudez-Edo, M., Elsaleh, T., Barnaghi, P., and Taylor, K.

(2016). Iot-lite: a lightweight semantic model for the

internet of things. In 2016 INTL IEEE conferences

on ubiquitous intelligence & computing, advanced

and trusted computing, scalable computing and com-

munications, cloud and big data computing, internet

of people, and smart world congress (uic/atc/scal-

com/cbdcom/iop/smartworld), pages 90–97. IEEE.

Cao, J., Carminati, B., Ferrari, E., and Tan, K.-L. (2009).

Acstream: Enforcing access control over data streams.

In 2009 IEEE 25th International Conference on Data

Engineering, pages 1495–1498. IEEE.

Carminati, B., Ferrari, E., Cao, J., and Tan, K. L. (2010).

A framework to enforce access control over data

streams. ACM Transactions on Information and Sys-

tem Security (TISSEC), 13(3):1–31.

Carminati, B., Ferrari, E., and Guglielmi, M. (2015). De-

tection of unspeciﬁed emergencies for controlled in-

formation sharing. IEEE Transactions on Dependable

and Secure Computing, 13(6):630–643.

Carminati, B., Ferrari, E., and Tan, K. L. (2007a). Enforcing

access control over data streams. In Proceedings of

the 12th ACM symposium on Access control models

and technologies, pages 21–30.

Carminati, B., Ferrari, E., and Tan, K. L. (2007b). Specify-

ing access control policies on data streams. In Interna-

tional Conference on Database Systems for Advanced

Applications, pages 410–421. Springer.

Celdr

an, A. H., Clemente, F. J. G., P

erez, M. G., and

erez, G. M. (2014). Secoman: A semantic-aware

policy framework for developing privacy-preserving

and context-aware smart applications. IEEE Systems

Journal, 10(3):1111–1124.

Chen, H., Finin, T., and Joshi, A. (2003). An ontology

for context-aware pervasive computing environments.

The knowledge engineering review, 18(3):197–207.

Choi, C., Choi, J., and Kim, P. (2014). Ontology-based

access control model for security policy reasoning in

cloud computing. The Journal of Supercomputing,

67(3):711–722.

Guo, H., Li, Y., and Jajodia, S. (2007). Chaining wa-

termarks for detecting malicious modiﬁcations to

streaming data. Inf. Sci., 177(1):281–298.

Home Assistant (2021). Open source home automation.

http://www.home-assistant.io/.

Janes, B., Crawford, H., and OConnor, T. (2020). Never

ending story: Authentication and access control de-

sign ﬂaws in shared iot devices. In 2020 IEEE Secu-

rity and Privacy Workshops (SPW), pages 104–109.

IEEE.

Jeong, Y.-S., Lee, S.-H., and Shin, S.-S. (2014). Access

control protocol based on privacy property of patient

in m-healthcare emergency. Wireless personal com-

munications, 79(4):2565–2578.

Kolozali, S., Bermudez-Edo, M., Puschmann, D., Ganz,

F., and Barnaghi, P. (2014). A knowledge-based ap-

proach for real-time iot data stream annotation and

processing. In 2014 IEEE International Conference

on Internet of Things (iThings), and IEEE Green Com-

puting and Communications (GreenCom) and IEEE

Cyber, Physical and Social Computing (CPSCom),

pages 215–222. IEEE.

Le-Phuoc, D., Dao-Tran, M., Parreira, J. X., and Hauswirth,

M. (2011). A native and adaptive approach for uniﬁed

processing of linked streams and linked data. In Inter-

national Semantic Web Conference, pages 370–388.

Springer.

Lee, C.-C., Hsu, C.-W., Lai, Y.-M., and Vasilakos, A.

(2013). An enhanced mobile-healthcare emergency

system based on extended chaotic maps. Journal of

medical systems, 37(5):9973.

Lu, R., Lin, X., and Shen, X. (2012). Spoc: A secure and

privacy-preserving opportunistic computing frame-

work for mobile-healthcare emergency. IEEE transac-

tions on parallel and distributed systems, 24(3):614–

624.

Manogaran, G., Thota, C., Lopez, D., and Sundarasekar, R.

(2017). Big data security intelligence for healthcare

industry 4.0. In Cybersecurity for Industry 4.0, pages

103–126. Springer.

Mengke, Y., Xiaoguang, Z., Jianqiu, Z., and Jianjian, X.

(2016). Challenges and solutions of information secu-

rity issues in the age of big data. China Communica-

tions, 13(3):193–202.

Montanari, R., Toninelli, A., and Bradshaw, J. M. (2005).

Context-based security management for multi-agent

systems. In IEEE 2nd Symposium on Multi-Agent Se-

curity and Survivability, 2005., pages 75–84. IEEE.

Nambiar, S., Kalambur, S., and Sitaram, D. (2020). Model-

ing access control on streaming data in apache storm.

Procedia Computer Science, 171:2734–2739.

Nehme, R. V., Lim, H.-S., and Bertino, E. (2013). Fence:

Continuous access control enforcement in dynamic

data stream environments. In Proceedings of the third

ACM conference on Data and application security and

privacy, pages 243–254.

Nehme, R. V., Rundensteiner, E. A., and Bertino, E. (2008).

A security punctuation framework for enforcing ac-

cess control on streaming data. In 2008 IEEE 24th

International Conference on Data Engineering, pages

406–415. IEEE.

Penrig, A., Song, D., and Tygar, D. (2000). Elk, a new

protocol for efﬁcient large-group key distribution. In

Proceedings 2001 IEEE Symposium on Security and

Privacy. S&P 2001, pages 247–262. IEEE.

Powell, J., Menon, D., and Jones, J. (1996). The effects of

hypoxaemia and recommendations for postoperative

oxygen therapy. Anaesthesia, 51(8):769–772.

Preuveneers, D., Van den Bergh, J., Wagelaar, D., Georges,

A., Rigole, P., Clerckx, T., Berbers, Y., Coninx, K.,

Jonckers, V., and De Bosschere, K. (2004). Towards

an extensible context ontology for ambient intelli-

gence. In European Symposium on Ambient Intelli-

gence, pages 148–159. Springer.

Puthal, D., Nepal, S., Ranjan, R., and Chen, J. (2015). A

dynamic key length based approach for real-time se-

curity veriﬁcation of big sensing data stream. In Inter-

Stream Data Access Control Model: The Need for Data Semantics

national conference on web information systems engi-

neering, pages 93–108. Springer.

Puthal, D., Wu, X., Nepal, S., Ranjan, R., and Chen, J.

(2017). Seen: A selective encryption method to en-

sure conﬁdentiality for big sensing data streams. IEEE

Transactions on Big Data.

Qin, W., Shi, Y., and Suo, Y. (2007). Ontology-based

context-aware middleware for smart spaces. Tsinghua

Science and Technology, 12(6):707–713.

Ren, J., Dubois, D. J., Choffnes, D., Mandalari, A. M.,

Kolcun, R., and Haddadi, H. (2019). Information

exposure from consumer iot devices: A multidimen-

sional, network-informed measurement approach. In

Proceedings of the Internet Measurement Conference,

pages 267–279.

Rhujittawiwat, T., Anderson, C., Keen, D., Miles, C.,

Farkas, C., Smiles, S., Wells, N., Roginski, J., Freder-

ick, S., and Banik, S. (2021). Making smart platforms

smarter: adding third party applications to home au-

tomation platforms. Journal of Computing Sciences

in Colleges, 37(5):43–53.

Sabelfeld, A. and Myers, A. C. (2003). Language-based

information-ﬂow security. IEEE Journal on selected

areas in communications, 21(1):5–19.

Sejdiu, B., Ismaili, F., and Ahmedi, L. (2021). Iotsas: an in-

tegrated system for real-time semantic annotation and

interpretation of iot sensor stream data. Computers,

10(10):127.

Shafagh, H., Burkhalter, L., Ratnasamy, S., and Hithnawi,

A. (2020). Droplet: Decentralized authorization and

access control for encrypted data streams. In 29th

USENIX Security Symposium (USENIX Security 20),

pages 2469–2486.

Stanford (2022). Prot

Thoma, C., Labrinidis, A., and Lee, A. J. (2019). Shoal:

Query optimization and operator placement for access

controlled stream processing systems. In IFIP An-

nual Conference on Data and Applications Security

and Privacy, pages 261–280. Springer.

Veltri, L., Cirani, S., Busanelli, S., and Ferrari, G. (2013).

A novel batch-based group key management protocol

applied to the internet of things. Ad Hoc Networks,

11(8):2724–2737.

Wang, X. H., Zhang, D. Q., Gu, T., and Pung, H. K. (2004).

Ontology based context modeling and reasoning using

owl. In IEEE annual conference on pervasive comput-

ing and communications workshops, 2004. Proceed-

ings of the second, pages 18–22. Ieee.

Whitehouse, K., Zhao, F., and Liu, J. (2006). Semantic

streams: A framework for composable semantic inter-

pretation of sensor data. In European Workshop on

Wireless Sensor Networks, pages 5–20. Springer.

Yu, W., Liu, Z., Chen, C., Yang, B., and Guan, X. (2017).

Privacy-preserving design for emergency response

scheduling system in medical social networks. Peer-

to-Peer Networking and Applications, 10(2):340–356.

Zhu, S., Setia, S., and Jajodia, S. (2006). Leap+ efﬁcient

security mechanisms for large-scale distributed sen-

sor networks. ACM Transactions on Sensor Networks

(TOSN), 2(4):500–528.

KMIS 2022 - 14th International Conference on Knowledge Management and Information Systems