Loading [MathJax]/jax/output/SVG/jax.js
Research article

Measurement data intrusion detection in industrial control systems based on unsupervised learning


  • Received: 23 September 2021 Accepted: 06 December 2021 Published: 10 December 2021
  • Anomaly detection strategies in industrial control systems mainly investigate the transmitting network traffic called network intrusion detection system. However, The measurement intrusion detection system inspects the sensors data integrated into the supervisory control and data acquisition center to find any abnormal behavior. An approach to detect anomalies in the measurement data is training supervised learning models that can learn to classify normal and abnormal data. But, a labeled dataset consisting of abnormal behavior, such as attacks, or malfunctions is extremely hard to achieve. Therefore, the unsupervised learning strategy that does not require labeled data for being trained can be helpful to tackle this problem. This study evaluates the performance of unsupervised learning strategies in anomaly detection using measurement data in control systems. The most accurate algorithms are selected to train unsupervised learning models, and the results show an accuracy of 98% in stealthy attack detection.

    Citation: Sohrab Mokhtari, Kang K Yen. Measurement data intrusion detection in industrial control systems based on unsupervised learning[J]. Applied Computing and Intelligence, 2021, 1(1): 61-74. doi: 10.3934/aci.2021004

    Related Papers:

    [1] Ekram E. Ali, Miguel Vivas-Cortez, Rabha M. El-Ashwah . New results about fuzzy γ-convex functions connected with the q-analogue multiplier-Noor integral operator. AIMS Mathematics, 2024, 9(3): 5451-5465. doi: 10.3934/math.2024263
    [2] Ebrahim Analouei Adegani, Davood Alimohammadi, Teodor Bulboacă, Nak Eun Cho, Mahmood Bidkham . On the logarithmic coefficients for some classes defined by subordination. AIMS Mathematics, 2023, 8(9): 21732-21745. doi: 10.3934/math.20231108
    [3] Wenzheng Hu, Jian Deng . Hankel determinants, Fekete-Szegö inequality, and estimates of initial coefficients for certain subclasses of analytic functions. AIMS Mathematics, 2024, 9(3): 6445-6467. doi: 10.3934/math.2024314
    [4] Shujaat Ali Shah, Ekram Elsayed Ali, Adriana Cătaș, Abeer M. Albalahi . On fuzzy differential subordination associated with q-difference operator. AIMS Mathematics, 2023, 8(3): 6642-6650. doi: 10.3934/math.2023336
    [5] Ahmad A. Abubaker, Khaled Matarneh, Mohammad Faisal Khan, Suha B. Al-Shaikh, Mustafa Kamal . Study of quantum calculus for a new subclass of q-starlike bi-univalent functions connected with vertical strip domain. AIMS Mathematics, 2024, 9(5): 11789-11804. doi: 10.3934/math.2024577
    [6] Ebrahim Analouei Adegani, Nak Eun Cho, Davood Alimohammadi, Ahmad Motamednezhad . Coefficient bounds for certain two subclasses of bi-univalent functions. AIMS Mathematics, 2021, 6(9): 9126-9137. doi: 10.3934/math.2021530
    [7] K. Saritha, K. Thilagavathi . Differential subordination, superordination results associated with Pascal distribution. AIMS Mathematics, 2023, 8(4): 7856-7864. doi: 10.3934/math.2023395
    [8] Jianhua Gong, Muhammad Ghaffar Khan, Hala Alaqad, Bilal Khan . Sharp inequalities for q-starlike functions associated with differential subordination and q-calculus. AIMS Mathematics, 2024, 9(10): 28421-28446. doi: 10.3934/math.20241379
    [9] Ebrahim Amini, Mojtaba Fardi, Shrideh Al-Omari, Rania Saadeh . Certain differential subordination results for univalent functions associated with q-Salagean operators. AIMS Mathematics, 2023, 8(7): 15892-15906. doi: 10.3934/math.2023811
    [10] G. Murugusundaramoorthy, K. Vijaya, K. R. Karthikeyan, Sheza M. El-Deeb, Jong-Suk Ro . Bi-univalent functions subordinated to a three leaf function induced by multiplicative calculus. AIMS Mathematics, 2024, 9(10): 26983-26999. doi: 10.3934/math.20241313
  • Anomaly detection strategies in industrial control systems mainly investigate the transmitting network traffic called network intrusion detection system. However, The measurement intrusion detection system inspects the sensors data integrated into the supervisory control and data acquisition center to find any abnormal behavior. An approach to detect anomalies in the measurement data is training supervised learning models that can learn to classify normal and abnormal data. But, a labeled dataset consisting of abnormal behavior, such as attacks, or malfunctions is extremely hard to achieve. Therefore, the unsupervised learning strategy that does not require labeled data for being trained can be helpful to tackle this problem. This study evaluates the performance of unsupervised learning strategies in anomaly detection using measurement data in control systems. The most accurate algorithms are selected to train unsupervised learning models, and the results show an accuracy of 98% in stealthy attack detection.



    One of the most fascinating areas of Complex analysis is the study of geometric characterization of univalent functions in the open unit disc U. Because of the challenging problem in studying the class S(11) of univalent functions in U as a whole, several subclasses of it emerged. The most studied of these are the classes CCV, SST, CCV(β) and SST(β)(0β<1) of convex functions, starlike functions and, convex and starlike functions of order β, respectively. Since the image domains of U plays a significant role in their geometric characterization, various subclasses of S(11) have been receiving attention in different directions and perspectives (see [4,6,20,23,28,32,33,36,37,38]). For this reason, Ma and Minda [17] gave a unified treatment of both SST and CCV. They considered the class Ψ of analytic univalent functions ψ(ς) with Reψ(ς)>0 and for which ψ(U) is symmetric with respect to the real axis and starlike with respect to ψ(0) such that ψ(0)>0. They initiated the following classes of functions that generalized and unified many renowned subclasses of S(11):

    SST(ψ)={fA:ςf(ς)f(ς)ψ(ς)}

    and

    CCV(ψ)={fA:(ςf(ς))f(ς)ψ(ς)},

    where A is the class of analytic functions f(ς) of the form

    f(ς)=ς+n=2δnςn. (1.1)

    If

    ψ(ς)=1+2π2(log1+ς1ς)2,

    then CCV(ψ)=UCV is the Goodman class of uniformly convex functions [8,30], which was later modified and examined by Kanas and Wisniowska [13,14]. Similarly, SThpl(s)=S(1(1ς)s), CVhpl(s)=C(1(1ς)s)(0<s1), are made-known by Kanas and Ebadian [15,16], respectively. These consist of functions fA such that ςf(ς)/f(ς) and (ςf(ς))/f(ς) lie in the domain bounded by the right branch of a hyperbola

    H(s)={σeiθC:σ=1(2cosθs)s,|θ|<πs2}.

    More special families of Ma and Minda classes can be found in [3,9,10,24,25,26,31,34,39].

    Recently, Kanas et al [18] introduced novel subclasses STL(s) and CVL(s) of SST and CCV, respectively. Geometrically, they consist of functions f(ς)A such that ςf(ς)/f(ς) and (ςf(ς))/f(ς) lie in the region bounded by the limaçon defined as

    Ls(U)={u+iv:[(u1)2+v2s4]2=4s2[(u1+s2)2+v2]},s[1,1]{0} (1.2)

    as shown in Figure 1 for different values of s. s=0.35,0.5,0.6,0.71,0.75 and 1 corresponds to blue, red, green, gray, yellow and black. Some novel properties of these classes were derived in [18].

    Figure 1.  Image representing Ls(U) for different values of s.

    Motivated by this present work and other aforementioned articles, the goal in this paper is to continue with the investigation of some interesting properties of the classes STL(s) and CVL(s). To this end, the sharp bounds of the Hankel determinant, subordination conditions as well as some radius results for these novel classes are investigated.

    To put our investigations in a clear perspective, some preliminaries and definitions are presented as follows:

    Denoted by W is the class of analytic functions

    w(ς)=n=1wnςn,ςU (2.1)

    such that w(0)=0 and |w(ς)|<1. These functions are known as Schwarz functions. If f(ς) and g(ς) are analytic functions in U, then f(ς) is subordinate to g(ς) (written as f(ς)g(ς)) if there exists a Schwarz function w(ς)W such that f(ς)=g(w(ς)),ςU.

    Janowski [12] introduced the class P(A,B),1B<A1 of functions p(ς) satisfying the subordination condition

    p(ς)1+Aς1+Bς,

    or equivalently, satisfying the inequality

    |p(ς)1ABr21B2r2|(AB)r1B2r2,|ς|r(0<r<1). (2.2)

    As a special cases, P(1,1)P and P(12β,1)P(β)(0β<1) are the classes of functions of positive real part and that whose real part is greater than β, respectively (see [7]).

    Definition 2.1. Noonan and Thomas [22] defined for q1,n1, the qth Hankel determinant of f(ς)S11 of the form (1.1) as follows:

    Hq(n)=|δnδn+1δn+q1δn+1δn+2δn+q2δn+q1δn+q2δn+2q2| (2.3)

    This determinant has been studied by many researchers. In particular Babalola [2] obtained the sharp bounds of H3(1) for the classes SST and CCV. By this definition, H3(1) is given as:

    H3(1)=|δ1δ2δ3δ2δ3δ4δ3δ4δ5|=δ3(δ2δ4δ23)δ4(δ4δ2δ3)+δ5(δ3δ22),δ1=1,

    and the by triangle inequality,

    |H3(1)||δ3||δ2δ4δ23|+|δ4||δ4δ2δ3|+|δ5||δ3δ22|. (2.4)

    Clearly, one can see that H2(1)=|δ3δ22| is a particular instance of the well-known Fekete Szeg¨o functional |δ3μδ22|, where μ is a real number.

    Definition 2.2. [18] Let p(ς)=1+n=1cnςn. Then pP(Ls) if and only if

    p(ς)(1+sς)2,0<s12,ςU,

    or equivalently, if p(ς) satisfies the inequality

    |p(ς)1|<1(1s)2.

    Demonstrated in [18], was the inclusion relation

    {wC:|w1|<1(1s)2}Ls(U){wC:|w1|<(1+s)21}. (2.5)

    It is worthy of note that the function Ls(ς)=(1+sς)2 is the analytic characterization of Ls(U) given by (1.2). Also, Ls(ς) is starlike and convex univalent in U for 0<s12. Furthermore, Ls(ς)P(β), where β=(1s)2,0<s12.

    Definition 2.3. Let fA. Then fSTL(s) if and only if

    ςf(ς)f(ς)P(Ls),0<s12.

    Also, fCVL(s) if and only if

    zf(ς)STL(s),0<s12.

    Moreover, the integral representation of functions fSTL(s) is given as

    f(ς)=ςexp(ς0p(t)1tdt),pP(Ls),

    while that of gCVL(s) is given as

    g(ς)=ς0f(t)tdt,fSTL(s).

    Furthermore, the extremal functions for each of the classes are given by

    Ψs,n(ς)=ςexp(z0Ls(tn)1tdt),Ψs,n(ς)STL(s)=ςexp(2snςn+s22nς2n),n=1,2,3,=ς+2snςn+1+(n+4)s22n2ς2n+1+. (2.6)

    and for Ks,n(ς)CVL(s),

    Ks,n(ς)=ς0Ψs,n(t)tdt,Ψs,n(ς)STL(s),n=1,2,3,=ς+2sn(n+1)ςn+1+. (2.7)

    Lemma 3.1. [1] If wW is of the form (2.1), then for a real number σ,

    |w2σw21|{σ,forσ1,1,for1σ1,σforσ1.

    When σ<1 or σ>1, equality holds if and only if w(ς)=ς or one of its rotations. If 1<σ<1, then equality holds if and only if w(ς)=ς2 or one of its rotations. Equality holds for σ=1 if and only if w(ς)=ς(x+ς)1+xς(0x1) or one of its rotations while for σ=1, equality holds if and only if w(ς)=ς(x+ς)1+xς(0x1) or one of its rotations.

    Also, the sharp upper bound above can be improved as follows when 1σ1 :

    |w2σw21|+(1+σ)|w1|21(1<σ0)

    and

    |w2σw21|+(1σ)|w1|21(0<σ<1).

    Lemma 3.2. [16] If wW is of the form (2.1), then for some complex numbers ξ and η such that |ξ|1 and |η|1,

    w2=ξ(1w21)

    and

    w3=(1w21)(1|ξ|2)ηw1(1w21)ξ2.

    Lemma 3.3. [19,Theorem 3.4h,p. 132] Let q(ς) be univalent in U and let θ and φ be analytic in a domain D containing q(U) with φ(ω)0, when ωq(U). Set Q(ς)=ςq(ς)φ(q(ς)),h(ς)=θ(q(ς))+Q(ς), and suppose that either

    (i) h(ς) is convex, or Q(ς) is starlike,

    (ii)

    Reςh(ς)Q(ς)=Re(θ(q(ς))φ(q(ς))+ςQ(ς)Q(ς))>0.

    If p(ς) is analytic in U with p(0)=q(0),p(U)D and

    θ(p(ς))+ςp(ς)φ(p(ς))θ(q(ς))+ςq(ς)φ(q(ς))=h(ς), (3.1)

    then p(ς)q(ς), and q(ς) is the best dominant in the sense that ptqt for all t.

    Lemma 3.4. [29] Let h(ς)=1+n=1cnςn,G(ς)=1+n=1dnςn and h(ς)G(ς). If G(ς) is univalent in U and G(U) is convex, then |cn||d1|, for all n1.

    Lemma 3.5. [11] Let wW. If |w(ς)| attains its maximum value on the circle |ς|=r at a point ς0U, then we have ς0w(ς0)=kw(ς0), where k1.

    Throughout this work f(ς) is taken to be of the form (1.1) while w(ς) is of the form (2.1). In the next sections, the main results are presented.

    In this section, we assume 0<s12. First, we establish a few auxiliary results whose applications will be needed hereafter.

    Lemma 4.1. Let fSTL(s). Then

    |δn|(2s)n1(n1)!,n2. (4.1)

    Proof. From the definition of fSTL(s), we have

    ςf(ς)f(ς)=p(ς),pP(Ls), (4.2)

    where

    p(ς):=1+c1ς+c2ς2+c3ς3+. (4.3)

    Comparing the coefficients of ςn in (4.2), it follows that

    (n1)δn=cn1+δ2cn2+δ3cn3++δn1c1. (4.4)

    It is obvious from Lemma 3.4 and the fact that Ls(ς) is convex for 0<s12 that

    |cn|2s,n1. (4.5)

    Using this result in (4.4), we obtain

    |δn|2sn1n1m=1|δm|,δ1=1,n2. (4.6)

    We need to show (4.1) by Mathematical induction. For this reason, assume (4.1) is true and proceed to prove

    |δn+1|(2s)n(n)!,n2.

    From (4.6),

    |δ2|=|c1|(2s)1,|δ3|2s2(1+|δ2|)(2s)22!,|δ4|2s2(1+|δ2|+|δ3|)(2s)33!,

    and finally,

    |δn+1|2sn(1+(2s)1+(2s)22!+(2s)33!++(2s)n1(n1)!)=(2s)nn!.

    Therefore,

    |δn+1|(2s)n(n)!,n2.

    Hence, by Mathematical induction, we have the desired result.

    In view of Theorem 4.1 and the definition of functions in CVL(s), we are led to the following result.

    Lemma 4.2. Let fCVL(s). Then

    |δn|(2s)n1(n)!,n2.

    Lemma 4.3. Let fSTL(s). Then

    |δ2δ4δ23|19s412.

    The bound 19s412 is sharp for the function

    Ψs,1(ς)=ς+2sς2+52s2ς3+73s3ς4+. (4.7)

    Proof. For fSTL(s),

    ςf(ς)f(ς)=(1+sw(ς))2, (4.8)

    where wW. Comparing coefficients of ς,ς2 and ς3 in (4.8), we arrive at

    δ2=2sw1,δ3=s(w2+52sw21)andδ4=23sw3+83s2w1w2+73s3w31. (4.9)

    By Lemma 3.1, we obtain

    |δ2δ4δ23|=4s23|w1(1w21)(1|ξ|2)η(1w21)(3w21)ξ2+14sw21ξ(1w21)1916s2w41|.

    Let x=w1,ξ=y with 0x1 and |y|1. Then the triangle inequality gives

    |δ2δ4δ23|F(x,|y|),

    where

    F(x,|y|)=4s23(x(1x2)(1|y|2)+(1x2)(3x2)|y|2+14sx2|y|(1x2)+1916s2x4),

    and

    F|y|=4s23(2(1x2)(3x2x)|y|+14sx2(1x2))>0.

    This means that F(x,|y|) is increasing on the interval [0,1]. So,

    F(x,|y|)4s23((1x2)(12+x2(s4))+1916s2x4):=F(x),

    where

    F(x)=2x(s+8)+19s2x3>0,

    which implies that F(x) is an increasing function of x on [0,1]. Consequently,

    |δ2δ4δ23|4s23F(1)=19s412.

    Lemma 4.4. Let fCVL(s). Then

    |δ2δ4δ23|s29.

    The bound s29 is sharp for the function

    Ks,2(ς)=ς+s3ς3+3s220ς5+. (4.10)

    Proof. From the definition of fCVL(s) and (4.9), it is easy to see that

    |δ2δ4δ23|=|16s2w1w3+19s3w21w219s2w2219s4w41|.

    The rest of the proof follows as in Theorem 4.3.

    Lemma 4.5. Let fSTL(s). Then

    |δ2δ3δ4|2s3.

    The bound 2s3 is best possible for the function

    Ψs,3(ς)=ς+23sς4+718s2ς7+. (4.11)

    Proof. From (4.9), a computation gives

    δ2δ3δ4=2s2w1(w276sw21)83s2w1(w2158sw21)23sw3.

    Employing Lemma 3.2, we write the expression for w3, and applying the triangle inequality together with Lemma 3.1, we obtain

    |δ2δ3δ4|14s23+2s3[(1x2)(1|y|2)+x(1x2)|y|2], (4.12)

    where we have taken w1=x,ξ=y with 0x1 and |y|1. Let H(x,|y|) represents the right side of (4.12). Then

    H(x,|y|)|y|=4s3(1x2)(1x)0.

    Thus,

    H(x,|y|)H(x,0):=H(x),

    where

    H(x)=14s2x3+2(1x2)s3andH(x)=2s3(7s2x).

    It is clear that H(x) attains its maximum value at x=7s2. Thus, H(x)H(7s2)=2s3. Consequently,

    |δ2δ3δ4|2s3.

    Lemma 4.6. Let fCVL(s). Then

    |δ2δ3δ4|s6.

    This bound cannot be improved since the function

    Ks,3(ς)=ς+16sς4+118s2ς7+ (4.13)

    attains the equality.

    Proof. Using the definition of fCVL(s) and (4.9), we find

    δ2δ3δ4=14s3w3113s2w1w216sw3.

    Let w1=x(0x<1) and ξ=y with |y|1. Then applying Lemma 3.2 and following the procedure of proof as in Theorem 4.5, we arrive at the desired result.

    Lemma 5.1. Let fSTL(s). Then for a real number μ,

    |δ3μδ22|{s2(58μ)2,forμ5s28s,s,for5s28sμ5s+28s,s2(8μ5)2,forμ5s+28s.

    It is asserted also that

    |δ3μδ22|+(μ5s28s)|δ2|2s,5s+28s<μ58

    and

    |δ3μδ22|(μ5s+28s)|δ2|2s,58<μ<5s+28s.

    These inequalities are sharp for the functions

    {¯λΨs,1(λς),forμ(,5s28s)(5s+28s,),¯λΨs,2(λς),for5s28sμ5s+28s,¯λPx(λς),forμ=5s28s,¯λQx(λς),forμ=5s+28s,

    where |λ|=1 and

    ςPx(ς)Px(ς)=Ls(ς(x+ς)1+xς),ςQx(ς)Qx(ς)=Ls(ς(x+ς)1+xς),0x1.

    Proof. From (4.9), we have

    |δ3δ22|=s|w2s(8μ5)2w21|.

    Then using Lemma 3.1 with σ=s(8μ5)2, we obtain the required result. For μ=1 in Theorem 5.1, we deduce the following sharp result.

    Corollary 5.1. Let fSTL(s). Then

    |δ3δ22|s.

    Lemma 5.2. Let fCVL(s). Then for a real number μ,

    |δ3μδ22|{s2(56μ)6,forμ5s26s,s3,for5s26sμ5s+26s,s2(6μ5)6,forμ5s+26s.

    It is asserted also that

    |δ3μδ22|+(μ5s26s)|δ2|2s3,5s26s<μ56

    and

    |δ3μδ22|(μ5s+26s)|δ2|2s3,56<μ<5s+26s.

    These inequalities are sharp for the functions

    {¯λKs,1(λς),forμ(,5s26s)(5s+26s,),¯λKs,2(λς),for5s26sμ5s+26s,¯λPx(λς),forμ=5s26s,¯λQx(λς),forμ=5s+26s,

    where |λ|=1 and

    (ςPx(ς))Px(ς)=Ls(ς(x+ς)1+xς),(ςQx(ς))Qx(ς)=Ls(ς(x+ς)1+xς),0x1.

    Proof. Using the definition of CVL(s) and (4.9), we get

    |δ3δ22|=s3|w2s(6μ5)2w21|.

    Then using Lemma 3.1 with σ=s(6μ5)2, we obtain the desired result. For μ=1 in Theorem 5.2, we deduce the following sharp result.

    Corollary 5.2. Let fCVL(s). Then

    |δ3δ22|s3.

    Theorem 5.3. Let fSTL(s). Then

    |H3(1)|s236(2s+1)(57s3+12s2+46s+34)

    Proof. The proof is immediate from (2.4), Lemma 4.1, Lemma 4.3, Lemma 4.5 and Corollary 5.1.

    Theorem 5.4. Let fCVL(s). Then

    |H3(1)|s2540(2s+1)(12s2+6s+33)

    Proof. The proof is straightforward from (2.4), Lemma 4.2, Lemma 4.4, Lemma 4.6 and Corollary 5.2.

    Theorem 6.1. The CVL(s)-radius for the class SST(β) (where β=(1s)2) is given by

    R1={12+2ss2+s44s3+6s24s+7,fors212,15,fors=212. (6.1)

    This radius is sharp for the functions given by

    f0(ς)={z(1z)2s(2s),fors212,z1z,fors=212. (6.2)

    Proof. Let fSST(β). Then

    ςf(ς)f(ς)P(β),β=(1s)2.

    This means that there exists wW such that

    ςf(ς)f(ς)=1+2(1β)w(ς)1w(ς),

    Let

    ςf(ς)f(ς)=p(ς). (6.3)

    Then by Schwarz lemma,

    |p(ς)1|2(1β)r1r. (6.4)

    It follows from logarithmic differentiation of (6.3) that

    |(ςf(ς))f(ς)1||p(ς)1|+|ςp(ς)p(ς)|.

    It is known from [27] that for pP(β),

    |ςp(ς)p(ς)|2(1β)r(1r)(1+(12β)r).

    Using this result together with (6.4), we write

    |(ςf(ς))f(ς)1|2r(1β)(2+(12β)r)(1r)(1+(12β)r).

    We need to show that

    |(ςf(ς))f(ς)1|1β.

    However, it holds if

    3(12β)r2+2(2+β)r10.

    Let T(r)=3(12β)r2+2(2+β)r1. Then T(0)=1<0 and T(1)=64β>0 such that T(0)T(1)<0. Thus, there exists R1[0,1] such that

    3(12β)R21+2(2+β)R11=0. (6.5)

    Therefore, 3(12β)r2+2(2+β)r10 for all r<R1 and R1 is the smallest roots of (6.5) given by (6.1).

    For sharpness, we consider the functions f0(ς) defined by (6.2). At the point z=R1, we have

    |(ςf0(ς))f0(ς)1|=1(1s)2.

    Theorem 6.2. Let fSTL(s). Then fCCV for all z in the disc |ς|<R2, where R2 is the positive roots of the equation

    s3x53s2x4+(3s2)sx3(13s2)x2(3s+2)x+1=0. (6.6)

    Proof. For fSTL(s),

    ςf(ς)f(ς)=(1+sw(ς))2,

    where wW. Therefore

    Re(ςf(ς))f(ς)Re(1+sw(ς))22r|w(ς)1+sw(ς)|(1sr)22r(1|w(ς)|2)(1r2)(1sr),

    where we have used the extension of Schwarz lemma (see [21]). Thus,

    Re(ςf(ς))f(ς)(1sr)3(1r2)2r(1sr)(1r2)>0

    if (1sr)3(1r2)2r>0. Let T(r)=(1sr)3(1r2)2r. Then T(0)=1>0 and T(1)=2<0 with T(0)T(1)<0. Therefore, there exists R2[0,1] such that (1sR2)3(1R22)2R2=0. Hence, (1sr)3(1r2)2r>0 holds for all r<R2, where R2 is the smallest positive roots of (6.6).

    Theorem 6.3. Let pP. Then pP(Ls) for all z in the disc

    |ς|<2ss22+2ss2,0<s22. (6.7)

    Proof. Let pP. Then

    |p(ς)1+r21r2|<2r1r2,r(0,1).

    We want to prove that

    |p(ς)1|1(1s)2.

    Now,

    |p(ς)1||p(ς)1+r21r2|+2r21r22r1r2+2r21r2=2r1r<1(1s)2

    if (6.7) is satisfied. To show that the radius cannot be improved, we consider the function

    p0(ς)=1+ς1ς.

    Then for ς=2ss22+2ss2,

    |p0(ς)1|=|2ς1ς|=2ss2,

    which shows that equality is attained for (6.7).

    Corollary 6.1. The STL(s)-radius and CVL(s)-radius for the classes of starlike and convex functions are given by (6.7).

    Theorem 7.1. If p(ς) is analytic in U with p(0)=1 and satisfies the condition

    Re(ςp(ς)p(ς))<2ss1, (7.1)

    or

    Re(ςp(ς)p(ς))>2ss+1, (7.2)

    then p(ς)(1+sς)2 for 1<s<0.

    Proof. Let p(ς) be defined by

    p(ς)=(1+sw(ς))2 (7.3)

    Clearly, w(ς) is analytic in U with w(0)=0. To prove our result, it is required to show that |w(ς)|<1 for all ςU. From (7.3), a simple calculation gives

    ςp(ς)p(ς)=2sςw(ς)1+sw(ς).

    Suppose there exists a point ς0U such that

    max|ς||ς0||w(ς)|=|w(ς0)|=1.

    Then by Lemma 3.5, w(ς0)=eiθ and ς0w(ς0)=kw(ς0). Thus,

    Re(ς0p(ς0)p(ς0))=Re(2sς0w(ς0)1+sw(ς0))=2k(1Re(11+seiθ))>2k(s1s)2ss1.

    This contradicts (7.1). Therefore, there exists no ς0U such that |w(ς0)|=1. Thus |w(ς)|<1 in U, so that p(ς)(1+sς)2 for 1<s<0.

    Similarly,

    Re(ς0p(ς0)p(ς0))=2k(1Re(11+seiθ))<2k(s1+s)2ss+1,

    which contradicts the assumption (7.2). Hence, the proof is completed.

    Following the discussion demonstrated by Sharma et al in [35] for Theorem 3, we present the following results.

    Theorem 7.2. Let 1<B<A1, 0<s12 and p(ς)=1+Aς1+Bς. Then pP(Ls) if and only if

    12s+s21A1B1+A1+B1+2ss2 (7.4)

    or, equivalently, if and only if

    A{2ss2+(1s)2B,forB(BA)0,2ss2+(1+2ss2)B,forB(BA)0. (7.5)

    Proof. The proof follows the techniques presented in [35,Theorem 3] For B=0,A=0 and B=A, we give the following consequences of Theorem 7.2.

    Corollary 7.1.

    (i) p(ς)=1+AςP(Ls)0<A2ss2.

    (ii) p(ς)=1/(1+Bς)P(Ls)(s22s)/(1+2ss2)B<0.

    (iii) p(ς)=(1+Aς)/(1Aς)P(Ls)0<A(2ss2)/(2+2ss2).

    Corollary 7.2. Let 1<B<A1 and consider

    ςf(ς)f(ς)=1+Aς1+Bςand(ςf(ς))f(ς)=1+Aς1+Bς.

    Then fSTL(s) and fCVL(s), respectively if and only if conditions (7.4) or (7.5) is satisfied

    Applying Corollary 7.1 along with the integral representation for the classes STL(s) and CVL(s), respectively, we present the following examples.

    Example 7.3.

    (i) For 0<A2ss2,

    f1(ς)=ςexp(Aς)STL(s)andf2(ς)=exp(Aς)1ACVL(s).

    (ii) For s22s1+2ss2B<0,

    f3(ς)=ς1+BςSTL(s)andf4(ς)=1Blog(1+Bς)CVL(s).

    (iii) For 0<A2ss22+2ss2,

    f5(ς)=ς(1Aς)2STL(s)andf6(ς)=ς1AςCVL(s).

    Theorem 8.1. Let 1B<A1,0<s12 and p(ς) be analytic in U with p(0)=1 such that

    1+ρςp(ς)1+Aς1+Bς(ρR{0},ςU). (8.1)

    If

    |ρ|AB2s(1s)(1|B|), (8.2)

    then pP(Ls).

    Proof. Let q(ς)=(1+sς)2,0<s12. Then q(ς) is convex univalent in U. Consider the functions ϕ(ω)=ρ and θ(ω)=1. These functions are both analytic in a domain containing q(U) with ϕ(ω)0. A computation shows that

    Q(ς)=ρςq(ς)=2ρsς(1+sς)andh(ς)=1+ρςq(ς)=1+2ρsς(1+sς).

    Further,

    ReςQ(ς)Q(ς)12s1s>0

    and

    Reςh(ς)Q(ς)=ReςQ(ς)Q(ς)>0.

    Using Lemma 3.3, the subordination condition

    1+ρςp(ς)1+ρςq(ς)

    implies p(ς)q(ς). To complete the proof, it suffices to prove that the circular disc 2.2 is contained in the region bounded by the curve h(eiθ)(θ[0,2π)). To this end, we must show that

    |h(eiθ)1AB1B2|AB1B2.

    Now,

    |h(eiθ)1AB1B2|=|2ρseiθ(1+seiθ)+B(AB)1B2|2s|ρ|(1s)|B|(AB)1|B|2.

    Thus,

    2s|ρ|(1s)|B|(AB)1|B|2AB1|B|2

    if (8.2) is satisfied.

    Theorem 8.2. Let 0<s12 and p(ς) be analytic in U with p(0)=1 such that

    1+ρςp(ς)(1+sς)2(ρR{0},ςU). (8.3)

    If

    |ρ|2+s2(1s), (8.4)

    then pP(Ls).

    Proof. Following the same arguments as in the proof of Theorem 8.1, we arrive at where to show that

    (1+sς)21+2ρsς(1+sς):=h(ς).

    To achieve this, it is enough to show that the domain bounded by the limaçon is inside the region bounded by the curve h(eiθ)(θ[0,2π)). As a result, we need to find ρ for which

    |h(eiθ)1|(1+s)21.

    Now,

    |h(eiθ)1|=2|ρ|s|1+seiθ|2|ρ|s(1s)2s+s2

    if

    |ρ|2+s2(1s). (8.5)

    Theorem 8.3. Let 1B<A1,0<s12 and p(ς) be analytic in U with p(0)=1 such that

    1+ρςp(ς)(1+sς)2(ρR{0},ςU).

    If

    |ρ|(2s+s2)(1+|B|)2AB, (8.6)

    then

    p(ς)1+Aς1+Bς.

    Proof. Let q(ς)=1+Aς1+Bς. We have that q(ς) is convex univalent in U. Therefore, following the method of proof in Theorem 8.1, we arrive at where to show that

    (1+sς)21+ρ(AB)ς(1+Bς)2:=h(ς).

    For this, we need to establish that the region bounded by the limaçon lies inside the domain bounded by the curve h(eiθ)(θ[0,2π). A simple observation of (2.5) suggests it suffices to show

    |h(eiθ)1|(1+s)21.

    Now,

    |h(eiθ)1|=|ρ|AB|1+Bς|2|ρ|AB(1+|B|)2.

    But

    AB(1+|B|)2(1+s)21

    provided (8.6) holds. This completes the proof.

    Theorem 8.4. Let 0<s12 and p(ς) be analytic in U with p(0)=1 such that

    1+ρςp(ς)p(ς)(1+sς)2(ρR{0},ςU). (8.7)

    If

    |ρ|(2s)(1+s)2, (8.8)

    then

    p(ς)(1+sς)2.

    Proof. Let q(ς)=(1+sς)2. Then q(ς) is convex univalent in U. The function ϕ(ω)=ρ/ω and θ(ω)=1 are analytic in the domain containing q(U). Set

    Q(ς)=ςq(ς)ϕ(q(ς))=ρςq(ς)q(ς)=2ρsς1+sς

    and

    h(ς)=θ(q(ς))+Q(ς)=1+ρςq(ς)q(ς)=1+2ρsς1+sς.

    Then

    Reςh(ς)Q(ς)=ReςQ(ς)Q(ς)>11+s>0.

    From Lemma 3.3, the differential subordination

    1+ρςp(ς)p(ς)1+ρςq(ς)q(ς),ςU

    implies p(ς)q(ς). To finalize the proof, we need to prove

    (1+sς)21+2ρsς1+sς, (8.9)

    which is equivalent to showing

    Ls(U)h(U).

    It is easy to see that the transformation h(ς)=1+ρςq(ς)q(ς) maps U onto the disc D(a,r), where

    a=1(1+2|ρ|)s21s2<1andr=2|ρ|s1s2.

    Therefore (8.9) holds if and only if

    11(1+2|ρ|)s21s2<2|ρ|s1s2,

    which implies

    |ρ|>(1+s)(2s)2.

    We choose to omit the proof of the next theorem since it follows the same argument as in Theorem 8.3.

    Theorem 8.5. Let 1B<A1, 0<s12 and p(ς) be analytic in U with p(0)=1 such that

    1+ρςp(ς)p(ς)(1+sς)2(ρR{0},ςU).

    If

    |ρ|(1+|A|)(1+|B|)(2s+s2)AB, (8.10)

    then

    p(ς)1+Aς1+BςςU.

    Remark 8.1.

    (ⅰ) If we put p(ς)=ςf(ς)/f(ς) and p(ς)=(ςf(ς))/f(ς) in Theorem 8.1-8.5, we obtain the conditions on ρ for which the respective subordination conditions (8.1), (8.3) and (8.7) imply fSTL(s) and fCVL(s).

    (ⅱ) We note that our condition 0<s12 cannot be relaxed in Theorem 8.1 and Theorem 8.2. Otherwise, starlikeness of Q will not be achieved. As such, the proof of the theorems will be extremely difficult to obtain via Lemma 3.3.

    The Ma and Minda classes of functions are the comprehensive generalization of the classes SST and CCV. These classes are vital in (GFT) because of their importance in science and technology. To this end, continuous studies of their subfamily, which are related to Limaçon domain were investigated. Coefficients bounds, Fekete Szeg¨o inequality as well as the upper bounds of the third Hankel determinants for these subclasses were derived. Finally, the techniques of differential subordination were also used to obtain some restrictions for which analytic functions belonged to these families. In addition, to have more new theorems under present examinations, new generalization and applications can be explored with some positive and novel outcomes in various fields of science, especially, in applied mathematics. These new surveys will be presented in future research work being processed by authors of the present paper.

    The authors would like to thank the Rector of COMSATS University Islamabad, Islamabad, Pakistan, for providing excellent research and academic environment. The fourth author is supported by UKM grant: GUP-2019-032.

    The authors declare that they have no conflicts of interest.



    [1] A. Abbaspour, S. Mokhtari, A. Sargolzaei, K. K. Yen, A survey on active fault-tolerant control systems, Electronics, 9 (2021), 1513. doi: 10.3390/electronics9091513 doi: 10.3390/electronics9091513
    [2] N. Sultana, N. Chilamkurti, W. Peng, R. Alhadad, Survey on SDN based network intrusion detection system using machine learning approaches, Peer-to-Peer Netw. Appl., 12 (2019), 493–501. doi: 10.1007/s12083-017-0630-0 doi: 10.1007/s12083-017-0630-0
    [3] S. Mokhtari, A. Abbaspour, K. K. Yen, A. Sargolzaei, A machine learning approach for anomaly detection in industrial control systems based on measurement data, Electronics, 10 (2021), 407. doi: 10.3390/electronics10040407 doi: 10.3390/electronics10040407
    [4] V. Chandola, A. Banerjee, V. Kumar, Anomaly detection: A survey, ACM comput. surv. (CSUR), 41 (2009), 1–58. doi: 10.1145/1541880.1541882 doi: 10.1145/1541880.1541882
    [5] K. Paridari, N. O'Mahony, A. Mady, R. Chabukswar, M. Boubekeur, H. Sandberg, A framework for attack-resilient industrial control systems: Attack detection and controller reconfiguration, P. IEEE, 106 (2017), 113–128. doi: 10.1109/JPROC.2017.2725482 doi: 10.1109/JPROC.2017.2725482
    [6] J. Goh, S. Adepu, M. Tan, Z. S. Lee, Anomaly detection in cyber physical systems using recurrent neural networks, 2017 IEEE 18th International Symposium on High Assurance Systems Engineering (HASE), (2017), 140–145. doi: 10.1109/HASE.2017.36 doi: 10.1109/HASE.2017.36
    [7] A. Javaid, Q. Niyaz, W. Sun, M. Alam, A deep learning approach for network intrusion detection system, Eai Endorsed Transactions on Security and Safety, 3 (2016). doi: 10.4108/eai.3-12-2015.2262516 doi: 10.4108/eai.3-12-2015.2262516
    [8] H. Choi, M. Kim, G. Lee, W. Kim, Unsupervised learning approach for network intrusion detection system using autoencoders, The Journal of Supercomputing, 75 (2019), 5597–5621. doi: 10.1007/s11227-019-02805-w doi: 10.1007/s11227-019-02805-w
    [9] M. Masson, Darktrace OT threat finds: Detecting an advanced ICS attack targeting an international airport, Aug., 2007. Available from: http://shorturl.at/nuJ19.
    [10] Y. Liu, Y. Mu, K. Chen, Y. Li, J. Guo, Daily activity feature selection in smart homes based on pearson correlation coefficient, Neural Process. Lett., (2020), 1–17. doi: 10.1007/s11063-019-10185-8 doi: 10.1007/s11063-019-10185-8
    [11] S. Bornelöv, J. Komorowski, Selection of significant features using Monte Carlo feature selection, Challenges in Computational Statistics and Data Mining, (2016), 25–38. doi: 10.1007/978-3-319-18781-5_2 doi: 10.1007/978-3-319-18781-5_2
    [12] Q. Guo, W. Wu, D. L. Massart, C. Boucon, S. De Jong, Feature selection in principal component analysis of analytical data, Chemometr. Intell. Lab., 61 (2002), 123–132. doi: 10.1016/S0169-7439(01)00203-9 doi: 10.1016/S0169-7439(01)00203-9
    [13] J. G. Dy, C. E. Brodley, Feature selection for unsupervised learning, J. Mach. Learn. Res., 5 (2004), 845–889. doi: 10.5555/1005332.1016787 doi: 10.5555/1005332.1016787
    [14] Y. Li, B. Lu, Z. Wu, Hierarchical fuzzy filter method for unsupervised feature selection, J. Intell. Fuzzy Syst., 18 (2007), 157–169. doi: 10.5555/1368376.1368381 doi: 10.5555/1368376.1368381
    [15] Z table website provides all required information for using Z-test. Aug., 2021. Available from: https://www.ztable.net.
    [16] F. T. Liu, K. M. Ting, Z. Zhou, Isolation forest, 2008 eighth ieee international conference on data mining, (2008), 413–422. doi: 10.1109/ICDM.2008.17 doi: 10.1109/ICDM.2008.17
    [17] M. Wu, C. Jermaine, Outlier detection by sampling with accuracy guarantees, Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining, (2006), 767–772. doi: 10.1145/1150402.1150501 doi: 10.1021/ie302042c
    [18] H. Ma, Y. Hu, H. Shi, Fault detection and identification based on the neighborhood standardized local outlier factor method, Ind. Eng. Chem. Res., 52 (2013), 2389–2402. doi: 10.1021/ie302042c doi: 10.1021/ie302042c
    [19] V. Kotu, B. Deshpande, Chapter 13 - Anomaly Detection, Data Science (Second Edition), (2019), 447–465. doi: 10.1016/B978-0-12-814761-0.00013-7 doi: 10.1016/B978-0-12-814761-0.00013-7
    [20] M. Ashrafuzzaman, S. Das, A. A. Jillepalli, Y. Chakhchoukh, F. T. Sheldon, Elliptic Envelope Based Detection of Stealthy False Data Injection Attacks in Smart Grid Control Systems, 2020 IEEE Symposium Series on Computational Intelligence (SSCI), (2020), 1131–1137. doi: 10.1109/SSCI47803.2020.9308523 doi: 10.1109/SSCI47803.2020.9308523
    [21] P. J. Rousseeuw, K. V. Driessen, A fast algorithm for the minimum covariance determinant estimator, Technometrics, 41 (1999), 212–223. doi: 10.1080/00401706.1999.10485670 doi: 10.1080/00401706.1999.10485670
    [22] P. C. Mahalanobis, On the generalized distance in statistics, National Institute of Science of India, 1936. doi: 10.1007/s13171-019-00164-5 doi: 10.1007/s13171-019-00164-5
    [23] M. Sokolova, G. Lapalme, A systematic analysis of performance measures for classification tasks, Inform. process. manag., 45 (2009), 427–437. doi: 10.1016/j.ipm.2009.03.002 doi: 10.1016/j.ipm.2009.03.002
    [24] N. V. Chawla, K. W. Bowyer, L. O. Hall, W. P. Kegelmeyer, SMOTE: synthetic minority over-sampling technique, J. artif. intell. res., 16 (2002), 321–357. doi: 10.5555/1622407.1622416 doi: 10.5555/1622407.1622416
    [25] H. Shin, W. Lee, J. Yun, H. Kim, HAI 1.0: HIL-based Augmented ICS Security Dataset, 13th USENIX Workshop on Cyber Security Experimentation and Test (CSET 20), 2020. doi: 10.5555/3485754.3485755 doi: 10.5555/3485754.3485755
    [26] S. Choi, HIL-based Augmented ICS (HAI) Security Dataset, The Affiliated Institute of ETRI, South Korea, 2020. Available from: https://github.com/icsdataset/hai.
    [27] H. Shin, W. Lee, J. Yun, H. Kim, Implementation of programmable CPS testbed for anomaly detection, 12th USENIX Workshop on Cyber Security Experimentation and Test (CSET 19), 2019. doi: 10.5555/3359012.3359014 doi: 10.5555/3359012.3359014
    [28] W. Hwang, J. Yun, J. Kim, H. Kim, Time-series aware precision and recall for anomaly detection: considering variety of detection result and addressing ambiguous labeling, Proceedings of the 28th ACM International Conference on Information and Knowledge Management, (2019), 2241–2244. doi: 10.1145/3357384.3358118 doi: 10.1145/3357384.3358118
  • This article has been cited by:

    1. Amina Riaz, Mohsan Raza, Muhammad Ahsan Binyamin, Afis Saliu, The second and third Hankel determinants for starlike and convex functions associated with Three-Leaf function, 2023, 9, 24058440, e12748, 10.1016/j.heliyon.2022.e12748
    2. Mridula Mundalia, S. Sivaprasad Kumar, On a subfamily of starlike functions related to hyperbolic cosine function, 2023, 0971-3611, 10.1007/s41478-023-00550-1
    3. Afis Saliu, Semiu Oladipupo Oladejo, On Lemniscate of Bernoulli of q-Janowski type, 2022, 2714-4704, 961, 10.46481/jnsps.2022.961
    4. Afis Saliu, Kanwal Jabeen, Isra Al-shbeil, Semiu Oladipupo Oladejo, Adriana Cătaş, Kwok-Pun Ho, Radius and Differential Subordination Results for Starlikeness Associated with Limaçon Class, 2022, 2022, 2314-8888, 1, 10.1155/2022/8264693
    5. Afis Saliu, Kanwal Jabeen, Isra Al-Shbeil, Najla Aloraini, Sarfraz Nawaz Malik, On q-Limaçon Functions, 2022, 14, 2073-8994, 2422, 10.3390/sym14112422
    6. Muhammad Ashfaq, Afis Saliu, Syed Zakar Hussain Bukhari, Issra Batool, Fekete-Szegö and Hankel inequalities related to the sine function, 2024, 32, 2769-0911, 10.1080/27690911.2024.2396832
    7. Afis Saliu, Kanwal Jabeen, V. Ravichandran, Differential subordination for certain strongly starlike functions, 2024, 73, 0009-725X, 1, 10.1007/s12215-023-00904-5
    8. Timilehin Gideon Shaba, Serkan Araci, Babatunde Olufemi Adebesin, Fairouz Tchier, Saira Zainab, Bilal Khan, Sharp Bounds of the Fekete–Szegö Problem and Second Hankel Determinant for Certain Bi-Univalent Functions Defined by a Novel q-Differential Operator Associated with q-Limaçon Domain, 2023, 7, 2504-3110, 506, 10.3390/fractalfract7070506
    9. Afis Saliu, Kanwal Jabeen, Qin Xin, Fairouz Tchier, Sarfraz Nawaz Malik, Radius Results for Certain Strongly Starlike Functions, 2023, 15, 2073-8994, 1124, 10.3390/sym15051124
    10. Sa’ud Al-Sa’di, Ayesha Siddiqa, Bushra Kanwal, Mohammed Ali Alamri, Saqib Hussain, Saima Noor, Mapping Properties of Associate Laguerre Polynomial in Symmetric Domains, 2024, 16, 2073-8994, 1545, 10.3390/sym16111545
    11. Afis Saliu, Kanwal Jabeen, Jianhua Gong, Sarfraz Nawaz Malik, Hala Alaqad, Starlikeness Associated with Certain Strongly Functions, 2025, 24058440, e41703, 10.1016/j.heliyon.2025.e41703
  • Reader Comments
  • © 2021 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(3956) PDF downloads(129) Cited by(4)

Figures and Tables

Figures(4)  /  Tables(1)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog