Serial or Simultaneous? Possible Attack Strategies with an Arsenal of
Attack Tools
Yahel Giat
a
and Irit Nowik
b
Department of Industrial Engineering, Jerusalem College of Technology, Havaad Haleumi 21, Jerusalem, Israel
Keywords:
Sequential, Parallel, Pooling, Cyber, Optimization.
Abstract:
In various fields such as medicine, management, cyber operations, and military strategy, the choice between
sequential and parallel strategies is pivotal in achieving objectives, be it maximizing the likelihood of success
or minimizing the time to victory. This study considers a hacker who attempts to destroy a rival system using
multiple attacking tools. It is assumed that the success probability of each attack tool to destroy the system
is equal and independent of the other tools. Execution of an attack is time consuming and it is assumed that
this attack time increases exponentially with the number of tools used simultaneously. We consider different
attaching schemes that vary in their design and balance between parallel and sequential steps. Our findings
indicate that when the attack time for a multi-tool attack is extremely short, the optimal solution will be a
purely simultaneous attack. Conversely, if the attack time approaches the total time required for a sequential
attack, then the optimal solution will be a purely sequential approach. In between these extremes, we discover
that a mixed strategy is optimal. Interestingly, our numerical analysis reveals that in these mixed cases, it is
consistently more advantageous to initiate a simultaneous attack and then complement it with a sequential one.
Moreover, we demonstrate that as the probability of success increases, the optimum tends towards a sequential
attack.
1 INTRODUCTION
In the realms of many disciplines such as medicine,
management, cyber and military operational art, the
choice between sequential and parallel strategies
plays a crucial role in achieving objectives, whether
its maximizing the probability of success or minimiz-
ing the time until victory. A sequential strategy in-
volves using the available tools one at a time, pro-
gressing to the next only if the previous one fails. In
contrast, a parallel strategy entails the employment of
many or all available tools simultaneously in a con-
certed effort.
Practical examples to this choice are abundant.
In the field of medicine, for example, an oncologist
may find that her patient’s tumor responds to only two
types of chemotherapies. Should she treat with both
simultaneously to maximize the chance of full tumor
lysis before the tumor develops defensive mutations?
Or perhaps, she should begin with monotherapy so
that in the event the tumor develops immunity to it,
she has another viable treatment for her patient. In
a
https://orcid.org/0000-0001-7296-8852
b
https://orcid.org/0000-0003-0697-8862
certain fields of medicine, a mix of simultaneous and
sequential strategies have been developed. For exam-
ple, physicians treating HIV carriers have recognized
as early as 1995 that a cocktail of drugs has better
results than sequential monotherapy, and when a spe-
cific cocktail is less effective then another cocktail is
used (Lu et al., 2018). Thus, both a simultaneous and
sequential approach are used.
This question has intrigued many thinkers in the
field of combat and military. For example, Soucy
(2018) studies General MacArthur’s sequential ap-
proach against the Japanese in the South West Pa-
cific during World War II, and compares it with Gen-
eral Schwarzkopfs parallel war against Iraq during
Desert Storm. His conclusion is that due to the
Unites State’s considerable military power, a parallel
war is advantageous and allows it to “quickly shatter
an enemy’s strategic and operational ability to resist
(Soucy, 2018, p. 3)”.
In this paper, we consider a hacker trying to attack
and destroy a system. The hacker has at its disposal
a pool of attack tools, whereas the attacked system
can (with a known probability) withstand an attack by
any one of the tools and develop defense mechanisms
Giat, Y. and Nowik, I.
Serial or Simultaneous? Possible Attack Strategies with an Arsenal of Attack Tools.
DOI: 10.5220/0012433800003639
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 13th International Conference on Operations Research and Enterprise Systems (ICORES 2024), pages 171-178
ISBN: 978-989-758-681-1; ISSN: 2184-4372
Proceedings Copyright © 2024 by SCITEPRESS Science and Technology Publications, Lda.
171
that allow it to be immune to that specific tool. This
setting gives rise to multiple attack strategies rang-
ing from fully sequential attacks (i.e., in each period a
single tool is employed until the system is destroyed),
to purely simultaneous attack in which all the attack
tools are used simultaneously. In between these two
extremes there are strategies that mix the sequential
and simultaneous approaches in varying degrees.
In our model, we assume that the probability of
success of each tool is equal and independent of the
other tools and show that if the attack time is inde-
pendent of the number of tools employed in the at-
tack, then the purely simultaneous strategy is optimal
to the hacker. In contrast, if attack time is the sum of
the individual attack times of the tools that it employs,
then the fully sequential strategy is optimal.
When attack time is increasing concavely with the
number of tools then strategies that mix between si-
multaneous and sequential attacks can be optimal, de-
pending on the total number of tools available to the
hacker and the probability of each tool to succeed.
While our model is quite simplistic in its nature,
it provides insights that may be useful whenever there
are multiple resources available. Staggering resources
in contrast to pooling them in a concerted effort is
of interest in a large array of applications, including
military, cyber, health and managerial.
2 LITERATURE REVIEW
Sequential versus simultaneous strategies have been
considered in a wide range of scientific fields. In
computer science, Andrad
´
ottir et al. (2017) studies a
model in which queues or servers may be pooled. It is
well known that pooling queues and servers is advan-
tageous when servers are not subject to failure. How-
ever, when servers could fail, then there is a tradeoff
between efficiency (queue length) and risk (i.e., the
probability that a system will be overcrowded). See
also Sunar et al. (2021) who considers a similar prob-
lem with different constraints. Similarly, Cui et al.
(2018) compares between simultaneous attacks and
sequential attacks on cyber-physical systems.
Sequential strategies are often likened to a cau-
tious approach and are widely used in the medical
field, where the progression of treatment options de-
pends on patient response. A notable example of se-
quential strategy in medicine is the management of
cancer. Oncologists often begin with less aggressive
treatments, such as radiation therapy or chemother-
apy, and only switch to more invasive procedures like
surgery if the initial methods do not yield the de-
sired results. This approach minimizes the imme-
diate risks and side effects while keeping more po-
tent interventions in reserve but in some cases may
be more susceptible to mutants. For example, with
metastatic breast cancer, international guidelines rec-
ommend using sequential monotherapy unless there
is rapid disease progression (Cardoso et al., 2014).
A systematic review comparing between combina-
tion (i.e., simultaneous) and sequential therapy found
that there was no difference in overall survival be-
tween the two groups but found that when drugs were
given one at a time there may be more time before
the tumors grew back again thereby achieving longer
progression-free survival (Dear et al., 2013). In con-
trast, for the initial treatment of hypertension Mac-
Donald et al. (2017) found that combination therapy
is superior to sequential monotherapy. Another inter-
esting example is the medical management of patients
that are HIV positive. It is a long standing consensus
that a combination therapy is superior to sequential
monotherapy in stopping these patients from acquir-
ing AIDS (Lu et al., 2018). Despite this established
clinical approach, it has been suggested that certain
HIV subpopulations that are resistant to multi-drug
treatment may benefit from a sequential monotreat-
ment approach (Phillips et al., 2003).
Sequential and parallel strategies also have a place
in the field of management, where the objective is of-
ten to maximize business performance or minimize
operational challenges (e.g, Thompson and Kwort-
nik Jr, 2008). The sequential approach is frequently
employed when tackling complex problems or imple-
menting organizational changes (Read et al., 2001).
Managers may choose to take one step at a time, eval-
uating the effectiveness of each action before pro-
ceeding to the next. For instance, when faced with
declining profits, a company may first focus on cost-
cutting measures, followed by rebranding, and subse-
quently, market expansion. This sequential strategy
allows for a more measured evaluation of each phase
of the plan. Parallel strategies in management are of-
ten associated with rapid and comprehensive changes.
For instance, during a business turnaround, a com-
pany on the brink of failure may implement a com-
bination of cost-cutting measures, diversification, re-
branding, and fundraising simultaneously to expedite
a recovery. This approach aims to address multiple
critical issues concurrently, potentially leading to a
quicker turnaround, but it also involves higher risk
and resource allocation Xiong et al. (2019).
Strategic planning is a framework used in various
sectors, including business, government, and the mil-
itary, with the primary goal of achieving long-term
objectives. For instance, during a military campaign,
a parallel approach might involve deploying all avail-
ICORES 2024 - 13th International Conference on Operations Research and Enterprise Systems
172
able resources, such as ground forces, air support, and
electronic warfare, in a coordinated effort to achieve a
swift and overwhelming victory. This approach seeks
to minimize the time until winning by overwhelming
the adversary. This approach is often evident in large-
scale offensives, where all available resources are de-
ployed simultaneously to overwhelm the enemy. The
“shock and awe” strategy, for example, aims to inca-
pacitate the adversary through a concentrated and si-
multaneous show of force (McNaughton, 2019). This
parallel approach can be highly effective in achieving
a swift victory but comes with higher risks and re-
source commitment. Conversely, a sequential military
strategy is often employed when precision and limited
collateral damage are essential. Surgical strikes con-
ducted by special forces units exemplify this approach
(Sasikumar, 2019).
The choice between sequential and parallel strate-
gies in medicine, management, and strategic plan-
ning depends on the specific objectives, risks, and re-
sources available. Each approach has its merits and
drawbacks, and a balanced combination of both may
be the most effective solution in many cases. Context-
specific analysis is essential to determine the optimal
strategy that will either maximize the winning proba-
bility or minimize the time until winning, depending
on the situation at hand.
3 THE BASIC MODEL
The environment that we present comprises two rival
entities. The attacking entity (“the hacker”) desires
to eliminate a system (“the system”) that the hacked
entity (“the hacked”) has acquired. To do so, the
hacker has at its disposal N distinct attack tools, each
of which can potentially eliminate the system. The
system has a lifespan of T periods after which it is
obsolete. Therefore, if all the attacks on it fail, it is
expected to survive exactly T periods.
The probability of each attack tool to successfully
eliminate the system is q and independently of the
other attack tools. In this case, we assume that the
system is eliminated at the end of the period. If the
attack tool fails to eliminate the system, it is assumed
that the attacked system was either prepared or has
developed defences against this tool and therefore this
attack tool is rendered useless.
In the basic model we assume that the attack time
is exactly one period regardless of the number of tools
employed. That is, an attack using one tool and at-
tack using k > 1 tools will each require one period to
set up. In the next section we relax this assumption.
Consider N strategies. Strategy Q
k
, k = 0, ..., N 1
follows the subsequent steps:
1. Initialization: Success:=False, Pool:= N attack
tools.
2. Serial Phase: For j := 1, ...k
(a) In each period, attack with a single attack tool
from the pool of tools.
(b) Remove this tool from the pool.
(c) If the attack is successful then Success:=True;
Go to step 4.
3. Simultaneous Phase: In period k + 1 attack simul-
taneously with N k (remaining) attack tools. If
successful Success:=True.
4. Output: Success.
Notice that strategy Q
k
, will last at most k + 1 periods
and that when k = 0 or k = N 1 the strategies are
either purely simultaneous (k = 0) or only sequential
(k = N 1).
In Figure 1 we depict the strategy Q
2
when N = 5.
The hacker attacks in the first period with a single at-
tack tool. If this attack has succeeded then the sys-
tem’s survival was one period. Otherwise, the hacker
attacks again with another single tool, and if this at-
tack has succeeded then the system’s survival was two
periods. Otherwise, the hacker goes all in with the
three remaining tools at its disposal. If this simulta-
neous attack has succeeded then the system’s survival
was 3 periods and if not then the system’s survival
was T periods.
?
n = 0


tool 1
-
success
prob = q
u
survival =1
?
fail
prob = 1q


tool 2
-
success
prob = q
u
survival = 2
?
fail
prob = 1q


tools
3,4,5
-
success
prob = 1(1q)
3
u
survival = 3
?
fail
prob = (1q)
3
u
survival = T
Figure 1: The Q
2
strategy in the basic model when N = 5
attacking tools are available.
Serial or Simultaneous? Possible Attack Strategies with an Arsenal of Attack Tools
173
Denote X as the random variable representing the
survival (i.e., length of life) of the attacked system.
The expected value of X comprises three elements.
First, it must survive the sequential phase. The sys-
tem is eliminated after j periods, 1 j k, if and
only if it survived the first j 1 periods, but was suc-
cessfully attacked in the jth period. The probability
for this to happen is (1q)
j1
q and when this hap-
pens the system survived exactly j periods. Thus, the
sequential phase contributes to the system’s expected
survival
k
j=1
j(1q)
j1
q. The firm survives exactly
k + 1 units of time if the sequential phase failed, but
the simultaneous phase (that comprises N k attack
tools) was successful. The probability of this is
(1q)
k
(1(1q)
Nk
) = (1q)
k
(1q)
N
and therefore the contribution of the simultaneous
phase to the expected survival is (k+1)
(1q)
k
(1
q)
N
. Finally, the system survives all the attacks with
probability (1 q)
N
in which case it survives T peri-
ods and therefore this contributesT (1q)
N
to the sys-
tem’s expected survival. Thus, the system’s expected
survival when strategy Q
k
is employed is given by:
E
Q
k
= q
k
j=1
j(1q)
j1
+ (k+1)
(1q)
k
(1q)
N
+ T (1q)
N
(1)
Lemma 1. The expected survival is increasing with
k.
Proof. To show that a higher k results with longer sur-
vival we must show that for each k<N1, it holds that
E
Q
k+1
E
Q
k
> 0. By (1),
E
Q
k+1
E
Q
k
= q(k + 1)(1q)
k
+
(k+2)(1q)
k+1
(1(1q)
Nk1
)
(k+1)(1q)
k
(1(1q)
Nk
). (2)
Expanding the latter expressions gives
E
Q
k+1
E
Q
k
= q(k + 1)(1q)
k
+
(k+2)(1q)
k+1
(k+2)(1q)
N
(k+1)(1q)
k
(k+1)(1q)
N
, (3)
which can be simplified to
E
Q
k+1
E
Q
k
= q(k + 1)(1q)
k
(1q)
N
+
(k+2)(1q)
k+1
(k+1)(1q)
k
. (4)
The first and fourth terms above can be combined to
(k+1)(1q)
k+1
and when added to the third term
results with (1q)
k+1
. Therefore, (4) simplifies to
E
Q
k+1
E
Q
k
= (1 q)
k+1
(1 q)
N
, (5)
which is positive since k + 1 < N.
Lemma 1 implies that the more the strategy is si-
multaneous (i.e, shorter sequential phase) the better
for the hacker. In fact, under the assumptions of the
basic model the best strategy is the pure simultane-
ous strategy, Q
0
. This result follows from the fact
that there is no probabilistic loss from attacking si-
multaneously (since the probabilities for success of
each attack tool are independent) whereas time-wise
a simultaneous approach is advantageous since more
attack tools are employed in a shorter time. In the
next section, we revisit this feature of the model con-
cerning the time that the hacker needs to execute a
simultaneous attack. In addition, we extend the menu
of strategies that we consider to strategies that begin
with a simultaneous attack employing a subset of the
tools followed by sequential attacks using the remain-
ing tools.
4 GENERALIZED MODEL
We consider two extensions to the model described in
the previous section.
4.1 Exponential Attack Time
The attack time for simultaneous cyber attacks may
not increase at a constant rate with the number of
attacking tools due to a concept known as “decreas-
ing marginal cost”. This idea comes from economics
and refers to the phenomenon where the cost of pro-
ducing each additional unit of a good or service de-
creases as the overall quantity increases. In the con-
text of cyber attacks, think of each attacking tool as a
unit. As you add more tools, there may be synergies
or efficiencies gained in the overall attack execution
process. For example, some tools may share com-
mon requirements or dependencies, and once those
are set up, adding more tools becomes quicker and
less resource-intensive.
Additionally, attackers may develop scripts or au-
tomated processes that can be reused across different
tools, further reducing the attack time for each addi-
tional tool. This is similar to economies of scale in
manufacturing, where producing more units leads to a
lower cost per unit. In line with these ideas we assume
that the attack time for a simultaneous attack with k
tools is k
α
, where α satisfies; 0 α 1. Note that
when α = 0, the attack time is constant in the number
of attacking tools, which is consistent with the basic
model (see Section 3), whereas α = 1 assumes that
the attack time is linear in the number of attackers
(i.e., the attack time increases at a constant rate with
the number of attacking tools). In Figure 2 we de-
ICORES 2024 - 13th International Conference on Operations Research and Enterprise Systems
174
pict the strategy Q
2
when N = 5. The hacker attacks
in the first period with a single attack tool. If this at-
tack has succeeded then the system’s survival was one
period. Otherwise, the hacker attacks again with an-
other single tool, and if this attack has succeeded then
the system’s survival was two periods. Otherwise, the
hacker goes all in with the three remaining tools at
its disposal. If this simultaneous attack has succeeded
then the system’s survival was 2 + 3
α
periods and if
not then the system’s survival was T periods.
?
n = 0


tool 1
-
success
prob = q
u
survival =1
?
fail
prob = 1q


tool 2
-
success
prob = q
u
survival = 2
?
fail
prob = 1q


tools
3,4,5
-
success
prob = 1(1q)
3
u
survival = 2 + 3
α
?
fail
prob = (1q)
3
u
survival = T
Figure 2: The Q
2
strategy in the Exponential model when
N = 5 attacking tools are available.
4.2 Simultaneous-Sequential Strategies
In Section 3 and 4.1 we considered a class of strate-
gies {Q
k
}, in which the hacker begins with a sequen-
tial approach and then delivers a final simultaneous
attack. We now extend our consideration to include a
class of strategies {M
k
}, with this order reversed.
Strategy M
k
, k = 0, ..., N 1 follows the subse-
quent steps:
1. Initialization: Success:=False, Pool:= N attack
tools.
2. Simultaneous Phase:
(a) For (N k )
α
periods attack simultaneously
with N k attack tools.
(b) Remove these tools from the pool.
(c) If the attack is successful then Success:=True;
Go to step 4.
3. Serial Phase: For j := 1, ...k
(a) In each period, attack with a single attack tool
from the pool of tools.
(b) Remove this tool from the pool.
(c) If the attack is successful then Success:=True;
Go to step 4.
4. Output: Success.
In Figure 3 we depict the strategy M
2
when N = 5.
The first round of attack lasts 3
α
periods and the
hacker attacks with three attack tools. If this attack
has succeeded then the system’s survival was 3
α
peri-
ods. Otherwise, the hacker attacks again with a single
tool (attack duration of one period), and if this attack
has succeeded then the system’s survival was 3
α
+ 1
periods. Otherwise, the hacker attacks with the last
available tool. If this attack has succeeded then the
system’s survival was 3
α
+ 2 periods and if not then
the system’s survival was T periods.
?
n = 0


tools
1,2,3
-
success
prob = 1(1q)
3
u
survival =3
α
?
fail
prob = (1q)
3


tool 4
-
success
prob = q
u
survival = 3
α
+ 1
?
fail
prob = 1q


tool 5
-
success
prob = q
u
survival = 3
α
+ 2
?
fail
prob = 1q
u
survival = T
Figure 3: The M
2
strategy in the Exponential model when
N = 5 attacking tools are available.
Accordingly, the system’s expected survival when
strategy Q
k
is employed is given by:
E
Q
k
= q
k
j=1
j(1q)
j1
+
(k+(N k)
α
)
(1q)
k
(1q)
N
+ T (1q)
N
. (6)
The system’s expected survival when strategy M
k
is
employed is given by:
E
M
k
= (N k)
α
1 (1q)
Nk
+ q
k
j=1
(Nk)
α
+ j
(1q)
Nk+ j1
+T (1q)
N
(7)
Serial or Simultaneous? Possible Attack Strategies with an Arsenal of Attack Tools
175
The first term is the contribution of the simultaneous
attack, the multiplication of the probability of N k
tools to succeed (1 (1q)
Nk
) with the survival if
the event happens ((N k)
α
). The second term is the
contribution of the sequential phase, where notice that
the probability for surviving exactly (Nk)
α
+ j peri-
ods, j = 1, ..., k, is q multiplied by the probability to
fail in the earlier rounds (1q)
Nk+ j1
. The third
term is the contribution to the survival if no attack
succeeds.
Proposition 1. The expected survival for each group
of strategies (E
Q
k
, E
M
k
) is
increasing with k when α = 0
decreasing with k when α = 1.
Proof. Consider first the Q
k
strategies. By (6), the
difference E
Q
k+1
E
Q
k
where k < N1 is
E
Q
k+1
E
Q
k
= q(k+1)(1q)
k
+
(1q)
N
(Nk)
α
1(Nk1)
α
(1q)
k
(k+(Nk )
α
)
+ (1q)
k+1
(k+1+(Nk1)
α
), (8)
This can be simplified to
E
Q
k+1
E
Q
k
= (Nk1)
α
(1q)
k+1
(1q)
N
+ (1(Nk)
α
)
(1q)
k
(1q)
N
. (9)
When α = 1 this can be further reduced to
q(Nk1)(1q)
k
,
which is negative since k < N1. When α = 0 this is
positive by Lemma 1.
We now consider the M
k
strategies. Using well-
known summation formula and algebraic manipula-
tion, (7) can be rewritten as
E
M
k
= (N k)
α
+
+ (1q)
N
1
q
1
(1 q)
k
1 kq
(Nk)
α
.
(10)
Therefore, the difference E
M
k+1
E
M
k
, k < N 1 is
E
M
k+1
E
M
k
= (Nk1)
α
(Nk)
α
+(1q)
N
1
q
1
(1q)
k+1
1(k+1)q
(Nk1)
α
(1q)
N
1
q
1
(1q)
k
1kq
(Nk)
α
(11)
This can be simplified to
E
M
k+1
E
M
k
= X(α)
1 (1q)
N
+ (1q)
Nk1
(1q)
N
(12)
where X(α) := (N k1)
α
(N k)
α
. When α =
0 then X(α) = 0 and therefore E
M
k+1
E
M
k
= (1
q)
Nk1
(1q)
N
> 0. Conversely, when α = 0 then
X(α) = 1 and E
M
k+1
E
M
k
= (1q)
Nk1
1 < 0.
Proposition 1 implies that for extreme values of α
the optimal strategy is also at the extreme. If α = 0 a
purely simultaneous strategy is optimal since it mini-
mizes the system’s survival, whereas when α = 1 the
purely sequential strategy is optimal. Recall the dis-
cussion following Lemma 1, when α = 0 the simul-
taneous approach has the advantage of gaining a high
probability of success in a short period of time. In
contrast, when α = 1, the attack time is additive and
therefore a sequential approach is better since it per-
mits early elimination of the system if an early attack
is successful.
In the next section we numerically analyze the at-
tack strategies when attack time is concave with the
number of tools employed in the attack, i.e, when
α (0, 1).
5 NUMERICAL ILLUSTRATION
We now consider numerically the case when the at-
tack time parameter α is intermediate. In (6) and (7),
the last term of the expected survival, T (1q)
N
, is
independent of the strategy and therefore in what fol-
lows we consider
ˆ
E
Q
k
:= E
Q
k
T (1q)
N
and
ˆ
E
M
k
:=
E
M
k
T (1q)
N
in lieu of E
Q
k
and E
M
k
, respectively.
That is,
ˆ
E
Q
k
and
ˆ
E
M
k
represent the variable compo-
nents of the expected survival. Throughout this sec-
tion we let N = 10 and α = 0.5. We plot
ˆ
E
Q
k
and
ˆ
E
M
k
and determine the optimal strategy when for different
values of q.
Figures 4, 5 and 6 describe the variable compo-
nents of the expected survival (
ˆ
E
Q
k
,
ˆ
E
M
k
) of the differ-
ent strategies when q = 0.1, 0.4 and 0.7, respectively.
These examples illustrates two phenomena that holds
more generally. First, when q is low, any M
k
strategy
is less or equal to its companion Q
k
strategy. How-
ever, as q increases the M
k
strategies increase com-
pared to the Q
k
strategies and when q is sufficiently
high the M
k
strategies are equal or above their com-
panion Q
k
strategies.
Second, as q increases the optimal strategy shifts
to the higher (i.e., larger k) strategies. The intuition to
this is that when q is high then a sequential approach
has an advantage since it is more likely to eliminate
the system within a single or few period (compared to
the simultaneous that increases the probability of suc-
cess but at the cost of more time). In contrast, when
ICORES 2024 - 13th International Conference on Operations Research and Enterprise Systems
176
6
-
0 3 6 9
Var. component of exp. survival
Strategy
k
1.5
2.0
2.5
3
3.5
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
....
...
...
...
....
...
...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
r
r
r
r
r
r
r
r
r
r
b
b
b
b
b
b
b
b
b
b
Optimum
6
r
ˆ
E
Q
k
b
ˆ
E
M
k
Figure 4: The variable component of the expected survival,
ˆ
E
Q
k
and
ˆ
E
M
k
, when N = 10, α = 0.5 and q = 0.1.
6
-
0 3 6 9
Var. component of exp. survival
Strategy
k
1.5
2.0
2.5
3
3.5
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
..
.
.
..
.
.
.
..
..
..
.
.
..
..
..
.
..
.
..
.
..
..
...
......
....
.....
.....
.....
....
.....
.....
....
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
r
r
r
r
r
r
r
r
r
r
b
b
b
b
b
b
b
b
b
b
Optimum
6
r
ˆ
E
Q
k
b
ˆ
E
M
k
Figure 5: The variable component of the expected survival,
ˆ
E
Q
k
and
ˆ
E
M
k
, when N = 10, α = 0.5 and q = 0.4.
6
-
0 3 6 9
Var. component of exp. survival
Strategy
k
1.5
2.0
2.5
3
3.5
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
..
.
.
.
.
.
.
..
..
..
..
..
.
.
.
.
..
.
..
.
..
.
..
.
..
..
..
...
..
..
..
...
....
....
...
....
............................................. .............................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
r
r
r
r
r
r
r
r r r
b
b
b
b
b
b
b
b
b
b
Optimum
?
r
ˆ
E
Q
k
b
ˆ
E
M
k
Figure 6: The variable component of the expected survival,
ˆ
E
Q
k
and
ˆ
E
M
k
, when N = 10, α = 0.5 and q = 0.7.
q is low the advantage of saving time (since α < 1)
is relatively higher since it is more likely that multi-
ple tools will have to be used anyways due to the low
probability of each one to eliminate the system.
While the Q
k
strategies are lower than the M
k
strategies when q is sufficiently high, it appears that
this does not happen when it counts, i.e., the optimal
M
k
strategy is always less or equal to its companion
Q
k
strategy. This can be gleaned from Table 1 that
describes the optimal strategy for different values of
Table 1: Optimal strategy for different values of α and q.
q α =0.1 α =0.3 α =0.5 α =0.7 α =0.9
0.1 Sim M
1
M
2
M
5
Seq
0.2 Sim M
1
M
4
M
7
Seq
0.3 M
1
M
3
M
6
M
8
Seq
0.4 M
2
M
5
M
7
Seq Seq
0.5 M
4
M
7
M
8
Seq Seq
0.6 M
6
M
7
Seq Seq Seq
0.7 M
7
M
8
Seq Seq Seq
0.8 M
8
Seq Seq Seq Seq
0.9 M
8
Seq Seq Seq Seq
Notes: Sim and Seq denote the purely simultaneous and
purely sequential strategies, respectively.
α and q (here, too, N = 10).
6 CONCLUSIONS
In this paper, we consider the question of whether
“putting all your eggs in one basket” is advisable or
not. This dilemma of pooling resources versus saving
assets for later needs is applicable to many fields of
operation. In our modelling, pooling resources does
not create an advantage or disadvantage in the prob-
ability of success, since we assume that the success
probability of each attack tool is identical and inde-
pendent. Instead, by pooling resources the hacker
trades off between waiting for the attack time to com-
plete before the success can be realized and the fact
that this attack time is shorter than a sequence of
single-tool attacks of the same number of tools.
Therefore, when the attack time of a multiple-tool
attack is very short, there is no advantage to the se-
quential approach and the optimal solution will be a
purely simultaneous attack. In contrast, if the attack
time is nearing the total time of a sequential attack,
then the optimal solution will be a purely sequential
attack. In between, we find that the optimal solution
is mixed, where we show numerically that it is al-
ways preferable to begin with a simultaneous attack
and then complement it with a sequential attack. Ad-
ditionally, we demonstrate that as the probability of
success increases the optimum leans more towards a
sequential attack.
While our observations from the numerical exam-
ple are not formally proved, it appears the formu-
las that we derive for the M
k
and Q
k
strategies can
be mathematically analyzed to qualify these observa-
tions. It is also of interest to examine how changing
the modeling assumption that the probability of suc-
cess is independent of the other probabilities will af-
fect the results. This is left to future study.
Serial or Simultaneous? Possible Attack Strategies with an Arsenal of Attack Tools
177
REFERENCES
Andrad
´
ottir, S., Ayhan, H., and Down, D. G. (2017). Re-
source pooling in the presence of failures: Efficiency
versus risk. European Journal of Operational Re-
search, 256(1):230–241.
Cardoso, F., Costa, A., Norton, L., Senkus, E., Aapro, M.,
Andre, F., Barrios, C., Bergh, J., Biganzoli, L., Black-
well, K. L., et al. (2014). Eso-esmo 2nd interna-
tional consensus guidelines for advanced breast can-
cer (abc2). The Breast, 23(5):489–502.
Cui, P., Zhu, P., Xun, P., and Shao, C. (2018). Robustness of
cyber-physical systems against simultaneous, sequen-
tial and composite attack. Electronics, 7(9):196.
Dear, R. F., McGeechan, K., Jenkins, M. C., Barratt, A.,
Tattersall, M. H., and Wilcken, N. (2013). Combina-
tion versus sequential single agent chemotherapy for
metastatic breast cancer. Cochrane Database of Sys-
tematic Reviews, (12).
Lu, D.-Y., Wu, H.-Y., Yarla, N. S., Xu, B., Ding, J., and
Lu, T.-R. (2018). Haart in hiv/aids treatments: fu-
ture trends. Infectious Disorders-Drug Targets (For-
merly Current Drug Targets-Infectious Disorders),
18(1):15–22.
MacDonald, T. M., Williams, B., Webb, D. J., Morant,
S., Caulfield, M., Cruickshank, J. K., Ford, I.,
Sever, P., Mackenzie, I. S., Padmanabhan, S., et al.
(2017). Combination therapy is superior to sequential
monotherapy for the initial treatment of hypertension:
a double-blind randomized controlled trial. Journal of
the American Heart Association, 6(11):e006986.
McNaughton, A. (2019). “parallel warfare” in conflicts with
limited political aims.
Phillips, A. N., Youle, M. S., Lampe, F., Johnson, M.,
Sabin, C. A., Lepri, A. C., and Loveday, C. (2003).
Theoretical rationale for the use of sequential single-
drug antiretroviral therapy for treatment of hiv infec-
tion. AIDS, 17(7):1009–1016.
Read, D., Antonides, G., Van den Ouden, L., and
Trienekens, H. (2001). Which is better: Simultane-
ous or sequential choice? Organizational Behavior
and Human Decision Processes, 84(1):54–70.
Sasikumar, K. (2019). India’s surgical strikes: Response to
strategic imperatives. The Round Table, 108(2):159–
174.
Soucy, R. (2018). Serial vs. parallel war: An airman’s view
of operational art. Technical Report ADA274440,
School of Advanced Military Studies, US Army Com-
mand and General Staff College, Leavenworth, KS.
Sunar, N., Tu, Y., and Ziya, S. (2021). Pooled vs. dedicated
queues when customers are delay-sensitive. Manage-
ment Science, 67(6):3785–3802.
Thompson, G. M. and Kwortnik Jr, R. J. (2008). Pooling
restaurant reservations to increase service efficiency.
Journal of Service Research, 10(4):335–346.
Xiong, G., Dong, X., Lu, H., and Shen, D. (2019). Re-
search progress of parallel control and management.
IEEE/CAA Journal of Automatica Sinica, 7(2):355–
367.
ICORES 2024 - 13th International Conference on Operations Research and Enterprise Systems
178