Printer Friendly

ENHDF: intrusion detection system for distribution networks in smart grid.

INTRODUCTION

Reconciliation of state-of-the-craftsmanship data and interchanges innovation (ICT), control, and figuring is a discriminating empowering influence to encourage matrix modernization and improvement for the current electric force frameworks. Throughout the evolution, development in the shrewd network improvement, the expected discriminating foundation is slowly presented to the general population, such that some piece of the frameworks, particularly the appropriation systems, including keen metering interchanges alongside controls of appropriated era and interest reactions at utilization destinations will possibly represent various security dangers. Progressed Metering Infrastructure (AMI) in the appropriation system is basically involved endpoint based home region systems (Hans), a framework based remote sensor systems (Wsns), and access point-based neighborhood/field territory systems (Nans and Fans) [1], [2]. As of late, a middleware structural engineering outline has been proposed to merge heterogeneous nature of administration/experience (Qos/Qoe) -situated shrewd network provisions, for example, range proficiency, force booking, and security assurance [3], [4]. Meanwhile, a few overviews and exercises intricately tended to various security issues as far as privacy, honesty, and accessibility (CIA), from aloof attacks to dynamic assaults [5] _ [14], for example, listening stealthily, sticking, altering, parodying, modifying, and different strike against the convention stacks of the OSI model; these assaults are predicted inescapable and nontrivial inside the connection to the digital physical shrewd network. Among which a few literary works have stressed the interrelationship between digital and physical securities [5], [6], [15]. A power grid is a complex system connecting a variety of electric power generators to customers through power transmission and distribution networks across a large geographical area, as illustrated in Figure 1 (adapted from National Security Telecommunications Advisory Committee (NSTAC) Information Assurance Task Force (IATF)).

[FIGURE 1 OMITTED]

Main Contribution The Paper:

ENHDF it is concentrated on attacks against state estimation utilizing DC force stream models. We show false information infusion assaults from the assaults point of view. We first show that it is feasible for the attacker to infuse noxious estimations that can sidestep existing methods for terrible estimation location. We then take a gander at two conceivable assault situations. In the first assault situation, the attacker is obliged to get to some particular meters because of, for instance, diverse physical assurance of the meters. In the second attack situation, the aggressor is constrained in the assets accessible to bargain meters. For both situations, we think about two conceivable assault objectives: irregular false information infusion assaults, in which the assault means to discover any attack vector as long as it can prompt a wrong estimation of state variables, and focused on false information infusion strike, in which the attacker plans to discover an assault vector that can infuse discretionary blunders into certain state variables. We indicate that the assault can methodically and proficiently build strike vectors for false information infusion attacks in both assault situations with both attack objectives.

ENHDF two realistic attack scenarios can be handled. They are

[check] The attacker is either constrained to some specific matters (due to the physical protection of the meters)limited in the resources required to compromise metering

[check] An attacker can systematically and efficiently construct attack vectors in both scenarios, affecting state estimation

[check] Demonstrate the success of these attacks through simulation using the IEEE 9-bus, 14-bus, 30-bus, 118-bus, and 300-bus systems

[check] And the results indicate that security protection of the electric power grid must be revisited.

DC State Estimation:

Here we present a common formulation of the state estimation problem when using a DC power flow model.

z = [H.sub.x] + e (1)

In [1], X = [([x.sub.1], [x.sub.2],...., [x.sub.n]).sup.T] Represents the true stated by the system that are to be estimated, Z = [([z.sub.1], [z.sub.2],...., [z.sub.m]).sup.T] Represents the sensor measurements, H is a m x n Jacobian matrix where Hx is a vector of m linear functions linking measurement to state, and e = e = [([e.sub.1], [e.sub.2],....., [e.sub.m]).sup.T] Represents the random error in measurements. The force system is viewed as discernible if there are sufficient estimations to make state estimation conceivable. There are numerous sensor arrangement calculations that can recognize the set of sensor estimations that guarantee discerns of a force system [17]. Ordinarily, there are a larger number of sensors in the force system than those required for discerns, i.e. m > n. The base set of estimations required to gauge the n state variables is regularly alluded to as a set of essential estimations or key estimations. The staying set of estimations is alluded to as excess estimations. The excess estimations are helpful in distinguishing awful sensor estimations [17]. Note that for DC state estimation, any set of n estimations whose comparing columns in H is directly free are sufficient to understand for then state variables and thus constitute a set of fundamental estimations. In other words, an independent linear equations are sufficient to solve for n variables. When m is greater than n, as is the typical case, state estimation involves solving an over-determined system of linear equations. It can be solved as a weighted least squares problem to arrive at the following estimators.

[??] = [([H.sup.T]WH).sup.-1] [H.sup.T] WZ (2)

Where W is an inclining lattice whose components are the estimated weights. It is normal to build a W in light of the reciprocals of the difference of estimation blunder. As pointed out in [1], as long as the sensor estimation failure is thought to be ordinarily circulated with zero mean, other regularly utilized estimation criteria, in particular, greatest probability basis and least fluctuation measure likewise prompt the estimator in [2].

Bad Measurement Detection:

Sensor estimations utilized for state estimation may be wrong in view of gadget mis-configuration; gadget disappointments, vindictive activities or different mistakes and can unfavorably influence the evaluation of state variables. In this manner, it is to a great degree profitable for force framework administrators to locate the vicinity of awful estimations and recognize them. Numerous plans for recognizing, distinguishing and redressing terrible estimations have been proposed [18], [17]. A typical methodology [18], [17] for locating the vicinity of awful information is by taking a gander at L2 -norm of estimation leftover which is defined as takes:

[parallel]z-H[??][parallel] (3)

In equation [3], [??] is the state estimate and z--H[??] is the measurement residual, which is the difference between the vector of observed measurements and estimated measurements. Intuitively, when observed measurements, z, contain bad data, the [L.sub.2]--norm of the measurement residual will be high. Thus, if the value of the expression in [3] is greater than a certain threshold t it is assumed that bad data is present. Assuming that all state variables are mutually independent and that the sensor errors follow a normal distribution, it can be shown that [[parallel]z--H[??][parallel].sup.2] follows a chi-squared distribution with v = m -n degrees of freedom [18]. Threshold value [tau] can be determined through a hypothesis test with a significance level [varies].

Attack Scenarios:

First attack scenario:

attacker is constrained to accessing some specific meters due to, for example, different physical protection of the meters

Second attack scenario:

attacker is limited in the resources required to compromise metering

* Two realistic attack goals

* Random false data injection attacks: attacker aims to find any attack vector as long as it can lead to a wrong estimation of state variables

* Targeted false data injection attacks: attacker aims to find an attack vector that can inject a specific error into certain state variables

Commonly used commonly used state estimation methods are given below:

* Maximum Likelihood (ML)

* Weighted Least Square (WLS)

* Minimum Variance criterion

False Data Injection Attacks:

False information infusion assaults on state estimation [1] are those in which an attacker3 controls the sensor estimations to impel a discretionary change in the assessed worth of state variables without being located by the awful estimation location calculation of the state estimator. In [1], Yao et al. display false information infusion assaults that can sidestep the awful estimation recognition calculation portrayed in Section II-A1. Here, we condense the essential assault standard, attack situations and objectives from [1].

Attack Principle:

Let a = [([a.sub.1], [a.sub.2], [a.sub.3],.... [a.sub.m]).sup.T] be an attack vector representing the malicious data added to the original measurement vectorz = [([z.sub.1], [z.sub.2],....., [z.sub.m]).sup.T]. Let [Z.sub.a] = z + a represent the resulting modified measurement vector. Let [[??].sub.bad] and [??] represent the estimates of x when using the manipulated measurements [Z.sub.a] and original measurements Z respectively. Then [[??].sub.bad] can be represented as [??] + c, where c is the estimation error introduced by the attacker. It is proved that if the adversary chooses the attack vector, a, to be equal to Hc, then resulting manipulated measurement Za = Z + a can pass the bad measurement detection algorithm and as long as the original measurement Z can pass it. To see this, consider the L2- norm of the measurement residual with manipulated data

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

It is also assumed that the adversary has the capability to manipulate sensors measurements, either by compromising the sensor or the communication between the sensor and the control center. However, this capability of the attacker is constrained as follows:

Situation I--Limited access to meters: The assailant is confined to getting to some particular meters because of diverse physical assurances of meters. Case in point, meters placed in substations with physical edge control may be much harder to get to than those placed in a bolted box outside of a building.

Scenario II--Limited assets accessible to bargain meters: The assaulter is constrained in the assets needed to trade off meters. Case in point, the assaulter just has assets to bargain up to k meters (out of every last one of meters). Because of the constrained assets, the assaulter might additionally need to minimize the amount of meters to be barging.

For both of the above scenarios, two attack goals are considered, namely, random false data injection and targeted false data injection. In random false data injection, the adversary aims to find any attack vector that injects arbitrary errors into the estimates of state variables. In targeted false data injection, the adversary aims to find an attack vector that injects specific errors into the estimates of specific state variables chosen by him. For targeted false data injection, two cases are considered:

Motivation:

Specifically, for a given H, we aim to identify a set of sensors and a set of state variables such that, when the measurements from the sensors in the chosen set are protected and when the values of state variables from the chosen set can be verified independently, then an adversary cannot find attack vectors that can inject false data without being detected. Furthermore, we would like to identify the smallest of such sets. There exists a set of sensors such that, when the measurements from those sensors are protected, an adversary cannot inject false data without being detected is evident from the results in [1]. For a given integer k, Figure 2 in [1111] shows the estimated success probability of an attacker in injecting false data without being detected when he picks k measurements at random to manipulate. This success probability was estimated using multiple trials of picking k measurements at random to manipulate. If the success probability of an attacker is less than 1 for a given k, it implies that there exist sets of m--k measurements such that when they are protected an attacker cannot inject false data without being detected.

For both of the above situations, two attack objectives are recognized, specifically, arbitrary false information infusion and focused on false information infusion. In irregular false information infusion, the foe means to discover any strike vector that infuses discretionary mistakes into the appraisals of state variables. In focused on false information infusion, the enemy plans to discover an attack vector that infuses particular mistakes into the evaluations of particular state variables picked by him. For focused on false information infusion, two cases are viewed as: obliged and unconstrained. In the obliged case, the enemy means to discover a strike vector that infuses particular slips into the evaluations of particular state variables yet does not dirty the appraisals of other state variables. This case speaks to circumstances where the control focus may have autonomous approaches to check the appraisals of certain state variables, and to maintain a strategic distance from recognition, the enemy would not like to dirty them. In the unconstrained case, the enemy has no such concerns in regards to dirtying other state variables. Systems to distinguish strike vectors for both of the above depicted strike objectives and in each of the above portrayed strike situations, and also the viability of those systems on the IEEE 9-transport, 14-transport, 30-transport, 118-transport and 300-transport test frameworks, are displayed in [1]. We allude the book fans to [1] for subtle elements.

Scenario-1: It is accepted that the length of the force line could be gotten to by the assaulter is L particular meters. Deliberately the assailant is the no one but individual can alter the estimations of these L meters. From this it is chosen that the aggressor can't basically pick any c and utilize a = Hc is the vector for strike. For those meters that cannot be gotten to by the assailant, the infused slips must stay 0.

Scientifically, let [I.sub._meter] = { i_1,i_2,.....,i_1} be the set of records of the k meters that the assaulter has admittance to. The assaulter can alter estimations Z_ij where ij[member of][I.sub.meter]. To launch a false information infusion attack without being caught, the assailant needs to discover a nonzero assault vector a=([varies]_1, [varies]_2,.... K_m)^t such that [varies]_1=0 for i [not member of] Imeter and a will be a straight consolidation of the segment vectors of H.

Random false Data Injection Attack:

In a random false data injection attack, the attacker aims to cause wrong estimation of state variables, where the errors injected into the wrong estimation could be any value. Thus the attacker vector a should satisfy the condition a = [([a.sub.1],[a.sub.2],...., [a.sub.m]).sup.T] = Hc with [a.sub.i] = 0for i [not member of] [I.sub.meter], where Imeter is the set of all indices of the meters which can be accessed by the attacker.

It is showed that c is redundant and can be eliminated from out formulation, and a = He can be transformed into an equivalent but more straight forward from which only has one variable a. This equivalent from will allow us to easily generate an attack vector a that satisfies the above condition.

Problem: a = He if and only if Ba = 0, where B = H[([H.sup.T]H).sup.-1] [H.sup.T]--I.

Proof: Let P = H[([H.sup.T]H).sup.-1] [H.sup.T] and B = P -I. According to Brock-well and Davis[1991], for any a[member of] [R.sup.m], Pa = a if and only if a is a linear combination of column vectors of H. [a = Hc]. Therefore,

A = He [??] Pa = a [??] Pa- a = 0 [??] Pa[P--I]a = 0 [??] Ba = 0. This means a satisfies a = He if and only if it satisfies Ba =0.

Generating a. The attacker needs to find a non-zero attack vector a such that Ba =0 and [a.sub.i] = 0for i [not member of] [I.sub.meter]. Represent a as a =[(0,..., [a.sub.i1],...0,...,0, [a.sub.i2],...0,[a.sub.ik], ...0, ...0).sup.T], where[a.sub.i1], [a.sub.i2], [a.sub.i3],.... [a.sub.ik] are unknown variables. Let B =([b.sub.1], [b.sub.2],....[b.sub.m]), where [b.sub.i] (1 [less than or equal to] i [less than or equal to]m) is the ith column vector of B. Thus, Ba = 0 [??] (...,[b.sub.i1], ..., [b.sub.i2],....,[b.sub.ik],.... )[(0, ...,0,[a.sub.i1],0, ...,0, [a.sub.i2] ,0,...,0, [a.sub.ik], 0, ...,0).sup.T] = 0. Let the m x k matrix B' = ([b.sub.i1], [b.sub.i2],....., [b.sub.ik]) and the length kveetora' = [([a.sub.i1],...., [a.sub.ik]).sup.T] . It becomes

Ba = 0 [??] B'a' (4)

On the off chance that the rank of B' is short of what k, B' is a rank-insufficient lattice, and there exist unbounded number of nonzero results a' that fulfill B'a' = 0 [meyer 2001]. As per Meyer [2001], the result is a' = (I--B'-b')d, where B'- is the Matrix 1-backwards of B' and d is a discretionary nonzero vector of length k. With a nonzero result a_, the assaulter can produce the attack vector a by filling 0s as the remaining components in a. On the off chance that the rank of B' is k, then B' is not a rank-lacking framework and B'a' = 0 has a novel result a' = 0 [meyer 2001]. This implies that no mistake might be infused into the state estimation, and the assault vector does not exist. As it were, the assaulter can't propel the attack.

If the rank of B's is the same as that of the augmented matrix (B's|y), B'sa' = y is a consistent equation, and there exist infinite solution a'=[B'.sup.-.sub.s] y + (I-[B'.sup.-.sub.s][B'.sub.s])d that satisfy B'sa'=y, where [B'.sup.-.sub.s] is the matrix 1-inverse of B's and d is an arbitrary nonzero vector of length k [Mayer 2001]. The attacker can generate an attack vector a from any a' [not equal to]0. If rank of B's is not the same as the rank of the augmented matrix (B's|y), then the relation B'sa' =y is not a consistent equation, and thus has no solution. This means that the attacker cannot generate an attack vector to inject the specific errors into the chosen state variables. How the attacker chooses specific errors cj for j[member of] [I.sub.variable] [h.sub.j] [c.sub.j] affects the feasibility of launching targeted attacks. Note that = [B.sub.s]b = [B.sub.s] [[summation].sub.jeIvatiable] [h.sub.j][c.sub.c]. If the attacker chooses [c.sub.j] such that [B.sub.s] [[summation].sub.jeIvatiable] [h.sub.j][c.sub.c] is a linear combination of columns of B's or [B.sub.s] [[summation].sub.jeIvatiable] [h.sub.j][c.sub.c] = 0, then the rank of the augmented matrix (B's|y) is the same as that of B's and the attacker can generate an attack vector. Otherwise, the attacker cannot generate an attack vector.

Scenario II--Limited Resources Available to Compromise Meters:

In Scenario II, we assume the attacker has resources to compromise up to k meters. Unlike Scenario I, there is no confinement on what meters could be picked. For the purpose of presentation, we call a length-m vector a k-meager vector on the off chance that it has at most k nonzero components. Subsequently, the aggressor needs to discover a k-scanty, nonzero assault vector athatfulfills the connection a = He.

Random False Data Injection Attack:

With the assets to trade off up to k meters, the attacker may utilize a beast energy methodology to develop an assault vector. That is, the aggressor may attempt all conceivable as comprising of k obscure components and m- k zero components. For every competitor a, the assaulter may check if there exists a nonzero result of such that Ba = 0 utilizing the same strategy as examined in Section 3.2.1. On the off chance that yes, the assaulter succeeds in developing an attack vector. Generally, the attacker has to attempt the following applicant. On the other hand, the beast power methodology could be drawn out. In the most pessimistic scenario, the assailant needs to inspect (m/k) applicant strike vectors. To enhance the time effectiveness, the assailant may exploit the accompanying perception. Since a fruitful strike vector is a direct consolidation of the section vectors of H (i.e., a = Hc), the assaulter can perform segment conversions to H to decrease the amount of nonzero components in the converted segment vectors. As this methodology proceeds, more segment vectors in the converted H will have fewer nonzero components. The section vectors with close to k nonzero components could be utilized as assault vectors. Specifically, when the matrix is a sparsematrix (which is typically the case in genuine force frameworks), it doesn't take numerous segment conversions to develop an attractive assault vector.

Heuristic Approach:

It is given a heuristic approach to exploit this observation. The attacker can initialize a size-n priority queue with the n column vectors of H. The attacker then repeats the following process: Take the column vector t with the minimum number of nonzero elements out of the queue. If t is a k-sparse vector, the algorithm returns and t can be used as the attack vector. If not, for each column vector s in the queue, the attacker checks if linearly combining t and s can result in a column vector with less zero elements than t. If yes, the attacker appends the resulting vector to the queue. The aggressor rehashes this methodology until a k-meager vector is discovered or the situated is unfilled. It is not difficult to see that a k-scanty vector built along these lines must be a direct mix of some segment vectors of H, and can serve as an attack vector. The heuristic methodology could be abate for a general H. On the other hand, it works pretty effectively for a sparse matrix h, which is generally the case for genuine force frameworks. Case in point, in our reenactment, when k = 4 in the IEEE 300-transport test framework, it takes the heuristic approach about 110ms on a customary PC to discover a strike vector. The heuristic methodology does not ensure the development of an attack vector even in the event that it exists, nor does it ensure the development of an attack vector that has the least number of nonzero components. In any case, it runs pretty immediately when it can build a strike vector, and in this manner could at present be a valuable instrument for the assailant.

Preferably, with a specific end goal to diminish the strike takes, the aggressor might want to bargain as few meters as could reasonably be expected. At the end of the day, the aggressor needs to discover the ideal strike vector with the base number of nonzero components. The assailant may utilize the animal energy methodology, talked about at the start of Section 3.3.1 with k being 1 at first; furthermore continuously expand k until a strike vector is found. Evidently, such a strike vector gives the ideal result with the base number of bargained meters. There are potential outcomes to enhance such a savage energy approach, for instance, by utilizing a double hunt in recognizing the base k.

Targeted False Data Injection Attack:

We take after the documentation utilized within Scenario-I to portray the focused on false information infusion attack. Let [I.sub.variabie] = {[i.sub.1], [i.sub.2],...,[i.sub.r]} wherer < n, indicate the set of files of the r target state variables picked by the agressor. In this assault, the assailant expects to build a strike vector a to supplant [[??].sub.i1] ..., and [[??].sub.i1], with [[??].sub.i1] + [c.sub.ir] respectively, where [c.sub.i1],...,[c.sub.ir] are the particular lapses to be infused. Like Scenario-I, we think about both obliged and unconstrained cases.

Common Principles:

Like false information infusion assaults, we consider a force framework comprising of m meters and n state variables for summed up false information infusion attacks. Review that the traded off estimations [Z.sub.a] might be spoken to as Za = Z + a, where z is the vector of unique estimations and an is the strike vector. [[??].sub.bad], the evaluated state variables got from [Z.sub.a] , could be spoken to as [??] + c, where c is the presented lapse and [??] is the genuine assessment. Hence, the 2-Norm of the estimation lingering of [z.sub.a] is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Let [tau] mean the discovery limit and [[tau].sub.a] = [tau]--[parallel]z--H[??][parallel]. On the off chance that[parallel]a--Hc[parallel] [less than or equal to] [[tau].sub.[alpha]], then [parallel][z.sub.a]--H[[??].sub.bad][parallel] [less than or equal to] [tau] and the agressor can sidestep the discovery. We allude to an attack in which the attack [parallel]a--HC[parallel] [less than or equal to] [tau] and the agressor can sidestep the discovery. We allude to an attack in which the attack vector a fulfills [parallel]a--Hc[parallel] [less than or equal to] [[tau].sub.a] as a summed up false information infusion assault. That is, in false information infusion strike, the assault vector ought to fulfill the condition[parallel]a--HC[parallel] = 0, while the summed up false information infusion attacks unwind this condition so that any vector a that fulfills [parallel]a--HC[parallel] [less than or equal to] [[tau].sub.a] might be used as the attack vector.

Scenario-I Limited Access to Meters:

Let [I.sub.meter] = {[i.sub.1],..., [i.sub.k]} speak to the set of lists of the k meters whose estimations could be traded off by the assailant. Accordingly, the aggressor can just change the estimation of the [i.sup.th.sub.j] meter to a wrong esteem, where [i.sub.j] [member of] [I.sub.meter].

RGFDIA

Expecta--He = t, where K is a length-mvector that reflects the contrast between a and Hc. The assailant cans sidestep identification as long as [parallel]t[parallel] = [parallel]a--Hc[parallel] [less than or equal to] [[tau].sub.a]. In irregular summed up false information infusion assaults, the vector c (i.e., the blunders acquainted with the state variables) might be any quality. Note that a might be spoken to as a = (0,...,0, [a.sub.i1],0, [a.sub.i2],0, [a.sub.ik], 0, ...,0)t, where [a.sub.il], [a.sub.i2] and [a.sub.ik] are the obscure variables to be dead set. Taking after Eqs. (5) and (6), we can get an equal type of the connection a--K = Heas follows:

a--K = Hc [??] B(a-t) = 0 [??] [B.sub.a] = [B.sub.t] [??] [B'.sub.a'] = Bt (5)

Where B = ([b.sub.1], ..., [b.sub.m]) = H[([H.sup.T]H).sup.-]1 [H.sub.T]-I,B' = ([b.sub.i1], ...,[b.sub.ik]),a = [([a.sub.i1], ...,[a.sub.ik]).sup.T], and K is a vector whose 2-Norm is less than [[tau].sub.a]. Thus, the attacker can solve a' from equation B'a' = Bt to get the attack vector a.

Tgfdia:

By propelling focused on summed up false information infusion strike, the assailant proposes to infuse particular failures into the estimation of picked state variables, while bringing about little residuals. We additionally consider both obliged and unconstrained cases. In the obliged case, the assaulter adjusts the target state variables however keeps the other state variables unaltered. Note that the presented failure c is an altered vector, and therefore the assaulter can specifically substitute c into a = Hc + t and conform t to get the assault vectora. Particularly, the assaulter can first utilize a zero vector as the beginningK = [([t.sub.1], ...,[t.sub.m]).sup.T]. Let f = [([f.sub.1], ...,[f.sub.m]).sup.T] For 1 [less than or equal to] i [less than or equal to] m, if [f.sub.i] [not equal to] and i [not member of] [I.sub.meter] then the assaulter can set [t.sub.i][t.sub.o]-[f.sub.i]. At long last, the assaulter checks whether the 2-Norm of the redesigned tis short of what [[tau].sub.a] or not. In the event that yes, the strike vector equivalents to Hc + t. Overall, the assault vector does not exist. In the unconstrained case, the aggressor alters the target state variables without any worry about the effect on the other state variables. This means just the components [c.sub.i] of c for i [member of] [I.sub.variable] are settled, and alternate components [c.sub.j] for j [not member of] [I.sub.variable] might be any qualities, where [I.sub.variable] = {[i.sub.1]..., [i.sub.r]}signify the set of files of the r target statevariables picked by the aggressor. Note that [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Let [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] = Hc might be converted into the following equivalent forms:

a-t = Hc [??] [B.sub.s] (a-t) = [B.sub.s]b [??] [B.sub.s]a = [B.sub.S] (t + b) [??] B's a' [??] [B.sub.S] (K + b),

whereBs = ([b.sub.s1] , ..., [b.sub.sm]) = [H.sub.s] [([H.sup.T.sub.s] [H.sub.s]).sup.-1] [H.sup.T.sub.s] - I,B's = ([b.sub.si1] ,...,[b.sub.sik]), a = [([a.sub.i1] ..., [a.sub.ik]).sup.T] and t is a vector whose 2-Norm is less than [[tau].sub.a]. Hence the attacker can solve a' from equation [B'.sub.s] a' = [B.sub.s]t + [B.sub.s]b to get the attack vector a.

Scenario-II Limited Access to Meters:

In Scenario II, the attacker can trade off up to k meters, yet there is no limitation on what meters could be traded off. The assailant needs to discover a k-inadequate, nonzero assault vector a that fulfills the bias[parallel]a Hc[parallel] [less than or equal to] [[tau].sub.a]. For arbitrary summed up false information infusion assaults and focused on summed up false information infusion attacks in the unconstrained case, the aggressor needs to discover a k- meager vector a that fulfills comparison [B.sub.a] = Bt and [B.sub.s]a = [B.sub.s](t + b), individually. The assaulter can straightforwardly diminish the issue to the Minimum weight Solution for Linear Equations issue by utilizing any vector with 2-Norm short of what or equivalent to [[tau].sub.a] ast. After the decrease, the attacker can exploit existing calculations, for example, Matching Pursuit [natarajan 1995; Pati et al. 1993; Lovisolo et al. 2005]; Basis Pursuit [chen 1995; Georgiev and Cichoki 2004, to discover a k--meager answer for mathematical statement Ba = Bt or [B.sub.s]a = [B.sub.s](t + b). For focused on summed up false information infusion strike in the obliged case, the attack vector an ought to be a k-inadequate vector that fulfills a = Hc + t. Note that the presented slip c is a settled vector. Consequently, if Hc is k-inadequate, then t = Oand a = Hc. Overall, accept that there are q (k < q [less than or equal to] m) nonzero components in Hc. The assaulter first modifies t such that q--k nonzero components in Hc could be drop when t is added to Hc, and afterward checks whether the 2-Norm of K is short of what or equivalent to [[tau].sub.a]. On the off chance that yes, Hc + t is an attack vector. Overall, the attack vector does not exist.

Simulation Results:

In this area we accept both unique and summed up false information infusion strike through tests utilizing IEEE test frameworks, including the IEEE 9-transport, 14-transport, 30-transport, 118-transport, and 300transport frameworks. The IEEE 9-transport, 14-transport, 30-transport, and 118-transport speak to partitions of American Electric Power System (in the Midwestern U.s.) in the early 1960s, while the IEEE 300-transport framework was produced by the IEEE Test Systems Task Force in 1993 [christie 1999]. In our analyses, we recreate assaults against state estimation utilizing the DC force stream model. We separate the arrangement of the IEEE test frameworks (especially grid H) from MATPOWER, a MATLAB bundle for tackling force stream issues [zimmerman and Murillo-sanchez 2007].2 We perform our analyses focused around network H and meter estimations got from MATPOWER. For each one test framework, the state variables are voltage plot of all transports, and the meter estimations are true force infusions of all transports and genuine force streams of all extension. For random false data injection attacks, k varied from 1 to the maximum number of meters in each test system. For each k, we randomly choose k specific meters to attempt an attack vector construction. We repeat this process 100 times for both IEEE 118-bus and 300-bus systems and 1,000 times for the other systems Estimate the success probability (probability of successfully constructing an attack vector with k given meters )denotes the percentage of the specific meters under the attacker's control, i.e. [k/total number of meters]

[p.sub.k] = Number of successful Trails/Number of Trails -(6)

[FIGURE 2 OMITTED]

The sample data and diagrams are taken from [123], and compared with this paper simulation results and shown in Figures.

Conclusion:

ENHDF, another class of ambushes, called false information infusion assaults was displayed, against state estimation in electric force frameworks. It is indicated that an assaulter can exploit the setup of a force framework to launch such ambushes to sidestep the current strategies for awful estimation identification. Two sensible strike situations: assailant is either obliged to some particular meters or constrained in the assets needed to trade off meters. Reenactments were performed on IEEE test frameworks to exhibit the achievement of these ambushes Results ENHDF demonstrate that the security insurance of the electric force framework must be revisited. In our future work, we might want to stretch out our results to state estimation utilizing Air conditioning force stream models. Additionally, we might likewise want to explore the likelihood of adjusting system aberrance discovery procedures to shield against false information infusion ambushes.

REFERENCES

[1.] Lo, C.-H. and N. Ansari, 2012. "The progressive smart grid system from both power and communications aspects," IEEE Commun. Surv. Tuts., 14(3): 799-821.

[2.] Lo, C.-H. and N. Ansari, 2013. "IEEE 802.15.4-based wireless sensor net-work design for smart grid communications," in Handbook on Green Information and Communications Systems, M. S. Obaidat, A. Anpala- gan, and I. Woungang, Eds. New York, NY, USA: Academic.

[3.] Zhou, L. and J. Rodrigues, 2013. "Service-oriented middleware for smart grid: Principle, infrastructure, and application," IEEE Commun. Mag., 51(1): 84-89.

[4.] Zhou, L., J.J.P.C. Rodriguesand L. Oliveira, 2012. "QoE-driven power scheduling in smart grid: Architecture, strategy, and methodology," IEEE Commun. Mag., 50(5): 136-141.

[5.] Sridhar, S., A. Hahn and M. Govindarasu, 2012. "Cyber_physical system security for the electric power grid," Proc. IEEE, 100(1): 210-224.

[6.] Mo, Y., T.-J. Kim, K. Brancik, D. Dickinson, H. Lee, A. Perrig and B. Sinopoli, 2012. "Cyber physical security of a smart grid infrastructure," Proc. IEEE, 100(1): 195-209.

[7.] Yan, Y., Y. Qian, H. Sharif and D. Tipper, 2012. "A survey on cyber security for smart grid communications," IEEE Commun. Surv. Tuts., 14(4): 998-1010.

[8.] Yang, Y., T. Littler, S. Sezer, K. McLaughlin and H. Wang, 2011. "Impact of cyber-security issues on smart grid," in Proc. 2nd IEEE PES ISGT Eur., pp: 1-7.

[9.] Liu, J., Y. Xiao, S. Li, W. Liang and C. Chen, 2012. "Cyber security and privacy issues in smart grids," IEEE Commun. Surv. Tuts., 14(4): 981-997.

[10.] Chen, X., K. Makki, K. Yen and N. Pissinou, 2009. "Sensor network security: A survey," IEEE Commun. Surv. Tuts., 11(2): 52-73.

[11.] Chen, P.-Y., S.-M. Cheng and K.-C. Chen, 2012. "Smart attacks in smart grid communication networks," IEEE Commun. Mag., 50(8): 24-29.

[12.] Berthier, R., W. Sanders and H. Khurana, 2010. "Intrusion detection for advanced metering infrastructures: Requirements and architectural directions," in Proc. 1st IEEE Int. Conf. Smart Grid Commun., pp: 350-355.

[13.] Sakarindr, P. and N. Ansari, 2007. "Security services in group communications over wireless infrastructure, mobile ad hoc, and wireless sensor networks," IEEE Wireless Commun., 14(5): 8-20.

[14.] Sakarindr, P. and N. Ansari, 2010. "Survey of security services on group communications," IET Inf. Security, 4(4): 258-272.

[15.] Neuman, C. and K. Tan, 2011. "Mediating cyber and physical threat propagation in secure smart grid architectures," in Proc. 2nd IEEE Int. Conf. Smart Grid Commun., pp: 238-243.

[16.] Fadlullah, Z., M. Fouda, N. Kato, X. Shen and Y. Nozaki, 2011. "An early warning system against malicious activities for smart grid communications," IEEE Netw., 25(5): 50-55.

[17.] Chen, X., K. Makki, K. Yen and N. Pissinou, 2009. "Sensor network security: A survey," IEEE Commun. Surv. Tuts., 11(2): 52-73.

[18.] Xiao, Z., Y. Xiao and D.-C. Du, 2013. "Exploring malicious meter inspection in neighborhood area smart grids," IEEE Trans. Smart Grid, 4(1): 214-226.

[19.] Liu, Y., M.K. Reiter and P. Ning, 2009. "False data injection attacks against state estimation in electric power grids," in CCS '09: Proceedings of the 16th ACM conference on Computer and communications security. New York, NY, USA: ACM, pp: 21-32.

[20.] http://www.ee.washington.edu/research/pstca/dyn30/dyn30dat.txt

(1) R. Balaambikha, (2) A. Thomas Paul Roy

(1) PG Scholar, Department of Computer Science and Engineering, PSNA College of Engineering and Technology, Dindigul, Tamil nadu, India.

(2) Assistant Professor, Department of Computer Science and Engineering, PSNA College of Engineering and Technology, Dindigul, Tamil nadu, India.,

Received 25 April 2016; Accepted 28 May 2016; Available 2 June 2016

Address For Correspondence:

R. Balaambikha, PG Scholar, Department of Computer Science and Engineering, PSNA College of Engineering and Technology, Dindigul, Tamilnadu, India.

E-mail: balaraja153@gmail.com.
Table 1: Timing Results For Scenario-I

                    Existing System

Test System         Random attack     Targeted Attack

IEEE 9 Bus          0.17-2.4          0.21-2.2
IEEE 14 Bus         0.16-5.6          0.26-11.3
IEEE 30 Bus         0.35-14.9         0.24-31.4
IEEE 118 Bus        0.34-867.9        0.42-1,874.5
IEEE 300 Bus        0.55-8,549.6      0.73-18,510

                    Proposed System

Test System         Random Attack     Targeted Attack

IEEE 9 Bus          0.16-2.2          0.12-1.87
IEEE 14 Bus         10.14-5.1         8.8-4.3
IEEE 30 Bus         0.32-12.34        0.28-10.21
IEEE 118 Bus        0.31-860.0        0.23-845.9
IEEE 300 Bus        0.48-8467.2       0.38-8342.1
COPYRIGHT 2016 American-Eurasian Network for Scientific Information
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Balaambikha, R.; Roy, A. Thomas Paul
Publication:Advances in Natural and Applied Sciences
Date:Jun 1, 2016
Words:6386
Previous Article:Six sense seamless handover protocol for 5G subscribers using artificial swarm intelligence.
Next Article:The effect of titanium oxide (Ti[O.sub.2]) addition on some properties of Unsaturated Polyester (UP).
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters