Protocol-based

Protocol Name

Cost of Communication

Cost of Memory

Its Advantages

Its Drawbacks

Detection Methodology

Key-usage based

Bloom Filter (PF) or called (Brooks et al. [12])

O (n log (n))

The high rate of false positives and negatives, and how to ensure that cloned nodes do not reliably communicate their credentials to the base station.

Centralized

Base station

based

SET [13]

O (n)

O (d)

Independent site, low

general memory

Single point of failure, Costly.

This protocol is very complex because its components are complex. The enemy can cancel the original nodes with this protocol [14].

Centralized

RED [15-19, 20]

O ($\sqrt{n}$)

O ($d \sqrt{n}$)

High detection ratio, low memory cost, unified

Uniform distribution of witnesses due to “pseudo-random selection of witnesses contract”

Deterministic, Need a trusted entity.

Centralized

CSI-1 [11, 21]

O (n log (n))

O (log (n))

High probability of detecting cloned nodes

Suffers from high communication and storage costs

Centralized

CSI-2 [21]

O (n)

O (1)

It has low communication and storage costs less than CSI-1.

The detection probability of clone nodes is low.

Centralized

Kenaza

et al. [22]

Summation of (BS*NS) to Trans & Response

Two keys.

A clone detection rate is high.

It is not scalable and has other common drawbacks of centralized solutions.

Centralized

PVM-MVP [23]

O (N²)

O (N)

Its accuracy is high in detecting cloned nodes.

Network throughput is higher due to lower network life, time consuming and cost-effective.

Centralized

Neighborhood social signature-based

Real-time detection Scheme (RTD) by: Xing et al. [24]

C. (1 + ratio)

O (d) + min (M,

w.log2 M)

Computational cost is low.

It could check itself for a fingerprint that matched the area in its vicinity because it couldn't handle an advanced clone [14].

Centralized

Cluster head

based

LNCA and Bloom Filter [6]

O (t²)

O (t)

Lower communication overhead

Lower detection probability.

Centralized

ABCD [25]

O (n log (n) )

O (n)

The detection rate of cloned nodes is high.

High communication cost, single point of failure, reduces network life-time.

Centralized

SWBC [26]

Communication overhead is reduced.

The detection rate is successful.

It needs a lot of storage space because the number of saved messages after detecting the enemy attack rate is greater when compared to RED and LSM.

Centralized

Zone based

ZBNRD [20]

O ($N \cdot \sqrt{n Z}$)+ O ($N z \cdot \sqrt{N}$)

O (d) / O (nZ)

The detection of cloned nodes is dynamic.

Be deterministic.

Centralized

Neighbor ID

based

X-RED [27]

O (n log (n))

O (n)

Higher detection probability, and low memory cost.

The traffic overhead is large.

Centralized

Node to

network

broadcasting

N2NB [28] or called Broadcast protocol (BP) [14]

O ($n^2$)

O (d)

Relatively acceptable communication cost if the network is small and provides a more efficient detection amount than centralized methods.

The larger the network, the greater the communication cost.

Distributed

Witness

node-based

DM [11, 28]

O (g log($\sqrt{n}$/ d )

O (g)

Lowers communication costs

Less security.

Distributed

RM [11, 20, 28]

O ($n^2$)

O ($\sqrt{n}$)

With enhanced flexibility, witnesses are difficult to predict.

It has lower detection probability, and high communication cost as well.

Distributed

LSM [11, 20, 28]

O ($n \sqrt{n}$)

O ($\sqrt{n}$)

- Better detection probability

- Improved communication cost

- It is distinguished from RM by lower connection cost and better memory efficiency.

He has two problems, the crowded center problem, and the crossover problem.

Distributed

LM [6]: (SDC & P-MPC [29, 30])

O (r.$\sqrt{n}$) + O(s)

O ($\omega$) Or: O(1) [30]

More efficient discovery than LSM and low memory

Its reliance on the trusted entity and the amount of cell size.

Communication cost (when the size of the cell is larger because if the size of the cell is smaller the node can be hacked easily).

Distributed

B-MEM [11, 29]

O (k.n. $\sqrt{n}$)

O (tk +$\mathrm{t}^{\prime} \mathrm{k} \sqrt{n}$)

Good detection probability, less memory

Location dependent.

Distributed

BC-MEM [11, 29]

O (tk + $\mathrm{t}^{\prime} \mathrm{k} \sqrt{n^{\prime}}$)

Solve the problems of crowded centers and crossover that LSM was suffering from. Less storage space, and a high probability of detection

Location dependent.

Distributed

C-MEM [11, 29]

O (t + $t^{\prime} \sqrt{n}$)

Location dependent.

Distributed

CC-MEM [11, 29]

O (t + $t^{\prime} \sqrt{n}$)

Location dependent.

Distributed

RDE [11, 31]

O (d.n. $\sqrt{n}$)

O (d)

Less memory consumption

It is not suitable for dynamic topological scenarios.

Distributed

Chano KIM [32]

O ($\sqrt{n}$)

O ($\sqrt{n}$)

The number decreases in contact messages.

No promising results in node cloning attack detection.

Distributed

RAWL [11, 19, 33]

O ($\sqrt{n}$ log (n)) or O (n$\sqrt{n}$) if a large network

O ($\sqrt{n}$ log(n)) or O($\sqrt{n}$) if a large network

High detection probability

Memory and communication costs are large.

Distributed

TRAWL [11, 19]

O ($\sqrt{n}$ log(n))

O $\left(1^2\right)^2$

High detection probability

High connection cost.

Distributed

NRDP [11, 17]

O (N.g($\sqrt{N}$)

O (g)

Share group membership information most simply.

Choosing a correspondent contract costs additional expenses.

Distributed

DHT [33, 34]

O (log n$\sqrt{n}$)

O (d)

Provides effective clone detection probability.

Higher communication cost.

Distributed

GDL and RMC [35]

O ($\sqrt{1} \times \sqrt{m} / 2$)

O ($\sqrt{n}$)

To provide a better level of use random validation and resilience.

Not resistant to smart node cloning attack due to determinism verification process.

Distributed

RWND [36]

Witness contract well insured, high probability of detection.

High connection, memory, and power cost when there are more regions.

Distributed

SSRWND [37]

Witness contract well insured, high probability of detection.

High connection, memory, and power cost when there are more regions.

Distributed

ERCD [38]

O (h$\sqrt{n}$)

O (h)

High detection probability with random witness selection.

Storing witnesses requires little routing of the node ring, which reduces storage costs.

Distributed

SBEA [39]

High complexity

Best

Improvement of the ERCD protocol by the added routing algorithm to give improvement to network lifetime and increase the performance of sensor nodes.

The arrival of data to the target is not guaranteed, The scattering density is high even if the active data node is far from the source and will not be affected by the resource-target node. Not useful if you need a contract.

Hybrid

PAWS [40]

O (3$\sqrt{n}$ log(n))

O (1)²

Detection probability, energy consumption, and flexibility are good.

Limited redundancy.

Distributed

RE-GSASA [41]

O (n$\sqrt{n}$)

O (n)

Improves the probability of detecting a clone attack by reducing the time to detect a clone attack and optimizing power consumption.

Messages overhead are high.

Distributed

ACTIVE [42]

O ($\sqrt{n}$)

O (1)

Good memory cost and the protocol tests nodes using relays to test whether randomly selected nodes are cloned.

Communication cost.

Distributed

Generation

or group

based

Bekara and Laurent [43, 44]

O (n)

O (1)

Simple protocol and its communication cost are lower.

The nodes in it are linked to their geographical locations and groups.

Distributed

Basic Protocol [45]

O (m)

O (m)

less communication,

Arithmetic operations and memory overhead are also lower.

It suffers from poor network connectivity so high-power applications cannot use it.

Distributed

Location Claim

Base Protocol [45]

O (m+d)

O (d+2m)

Strong detection ability, appropriate communication cost, and lower computational and storage burdens.

Dumping bogus claims due to DoS risks.

Distributed

Multi-Group

Base Protocol [45]

3xO(m + d)

O (d + 2xm(1 + Dmax)

A solid compromise to knots since the opponent wants to settle several sets.

High connection cost.

Distributed

Sei and Honiden [17]

O (r)

O (r.$\sqrt{n}$)

There is no entity more reliable, and more flexible.

large communication costs,

The start time of the built-in detection.

Distributed

CINORA [46]

O (M*$\sqrt{W}$) thus involved is O(M*W)

It does not conclusively determine the identity of a node to a cell, which helps us to make node anonymous for increased security.

The effect on detection accuracy is unknown when adversary nodes form a community to exploit the network and disrupt the communication process, and the Effect of symmetric and asymmetric switches on communication and memory overheads also.

Distributed

Neighbor based

HIP/HOP [42, 47]

Low communication cost

High Memory cost.

Distributed

Clustering based

LEACH [48, 49]

1. LEACH schedule-based protocols addressing idle listening avoidance using TDMA schemes.

2. Explicitly define the sending and receiving opportunities of the nodes and allow them to rest when they are not needed.

3. It calculates transmission schedules in such a way that no collisions occur at receivers.

4. There is therefore no need to use special mechanisms to avoid hidden end positions.

1. When faced with a changing topology, setting and maintaining tables requires signaling.

2. The time is divided into relatively small slots if the TDMA variant is used, and the sender and receiver must agree on the limits of the slots to meet each other and also avoid interference with other slots, which may lead to collisions.

However, solving the time synchronization problem involves some additional signaling traffic.

3. Despite the small date ranges, these schedules do not easily adapt to different pregnancies.

4. A node has difficulty giving unused bits of time to its neighbors.

5. The schedule of the node needs a large amount of memory and the nodes with weak capabilities cannot handle it.

Distributed

NI LEACH [50]

O (l (1+m’2))

O (k. e)

Less delay, balanced productivity.

It is not useful to use it in the case of multiple opponents, because its detection rate is lower.

Distributed

MMF - CND using LEACH Protocol [51]

CND: O(n)

MMF: O ($c^2$n)

CND O (n)

Add new security to LEACH protocol by MMF algorithm and node clone detection by CND algorithm.

Communication and memory costs are high.

Centralized

Hierarchical Distributed Algorithm (HDA) [14]

O ($N^2$)

O (N)

Transmitter by sensor nodes only to its cluster headers. Then the cluster head sends the data to the base station.

High communication and memory cost.

Distributed

NBCAD [42, 52]

Less than SET, RED Protocols

High Detection ratio and higher probability. Less memory and power consumption of nodes as well as reduced transmission range.

Distributed

Witness path

based

LSCD [53]

O (n$\sqrt{n}$)

O (l/ r )

High probability of detection, and node storage is suitably low.

The communication cost is

High.

Distributed

TAWS [6]

O ($\sqrt{n}$log(n))

O ($1^2$)

The  witness nodes are selection of in a random manner using a random walk table.

Relatively high memory cost due to the random walk table.

Distributed

Cluster head

based

LTBRD [54]

Less memory and power consumption.

The detection probability is low.

Distributed

PRCD [55]

O (Np)

O (1 / p)

Long network life and adequate computing complexity.

Distributed

Base station Based

DNCA

(Detection of Node Capture Attack) [56]

O (n$\sqrt{n}$)

O (n)

The captured nodes did not participate during this period in any network operation.

The communication and memory costs are high.

Distributed

Type of Protocol

The Name

Communication Cost

Memory

Cost

Advantages

Drawbacks

Detection Methodology

Node speed based [11]

Fast [14] by: Ho et al. Scheme [11]

O($n \sqrt{n}$)

O(n)

A mobile node should never move at speeds greater than the maximum speed configured by the system.

This protocol does not carry the costs of the current generation of Wireless Sensor Networks because it uses much more expensive equipment called GPS [14].

Centralized

Information exchange based [11]

or

Conflict based [14]

Witness based detection probability, time-based [42]

XED [14]

O(1)

O(4 . v . E(X))

Low memory cost because sensor node location information is not needed, only persistent communication is required [42].

If cloned nodes communicate with each other, they can create secret channels and can easily fool detection technology [14].

Distributed

Zhu et al. [57] by Token Based

-

-

Used two algorithms: token and seen many time based.

Clone nodes can share tokens and make the protocol exist in name only.

This protocol fails when a clever attacker creates secret channels between cloned nodes.

Distributed

Node meeting based [11] or

Node mobility [14, 39]

NBDS [14]

O(r$\sqrt{n}$)

O(r)

Independent of location.

The cost of messages is high.

Distributed

EDD Protocol [14]

O(1) / O(n)

O(n)

Composed of two phases: the offline phase and the online phase.

Inapplicable due to the high storage overhead for large-scale WSNs [11, 14].

Distributed

SDD/CDD by Conti et al. [58]

Both proposed algorithms are based on the simple observation that "If node a does not return node b within a certain period, node b may have been captured".

Any sensor node can flood the entire mobile WSN with a broadcast message which is not possible in reality. There is no change in membership in the network and this is not the case in reality.

SEDD [11]

O(n)

O($\xi$)

Instead of monitoring all nodes in EDD, each node only monitors a subset of nodes

Memory and communication cost is high stilled but it is better than EDD.

Distributed

Mobility assisted based [11]

or

Time location-based

PDRA [59] Wang and Shi Protocol with Base Station [11, 14]

O(n)

If the answers of the cloned nodes are collected by different patrol nodes, they will be discovered by exchanging messages with the patrol guards after the round, or by the base station.

Distributed

PDRA [59] Wang and Shi Protocol Without Base Station [11, 14]

O($n * \sqrt{u}$)

If cloned nodes are deployed in an area where a periodic node collects its answer message in a periodic period, the sentinel can invalidate it immediately upon receiving the second answer and overstepping the distance between the two locations.

Distributed

UTLSE [11, 14]

O(n)

O($\sqrt{n}$)

- Each node is configured with a unique tracking group.

- Statements of claims are only exchanged when appropriate witnesses meet each other.

High communication cost.

Distributed

MTLSD [11]

O(n)

O($\sqrt{n}$)

- Statements of claims are only exchanged when appropriate witnesses have met each other.

- MTLSD has a higher detection power than UTLSE.

High communication cost.

Distributed

Neighbor based

SHD [14]

It uses the fingerprint for each node, as well as the decision of the witness node.

High communication cost.

Distributed

Key usage-based

Deng and Xiong Protocol [11]

O(n log n)

Polynomial-based pair-wise key pre-distribution and Bloom Filters.

There is no evidence that a participating cloned node will faithfully report its keys to the base station.

The number of the parent node of pair switches may exceed a threshold value due to their connectivity.

The dependence of centralized detection protocol operations on base station participation leads to singlepoint failure and rapid depletion of power for sensor nodes around the base station.

Centralized

Detection rate

Detection probability [42]

CLONE WARS [42, 47]

High detection rate.

Communication cost is reduced.

Distributed

Type of Protocol

Name

Communication Cost

Memory

Cost

Advantages

Drawbacks

Detection Methodology

Topology distortion

MDS [11]

affordable

O(1)

High detection rate.

Communication overhead on the network in the case of dense network topologies.

Hybrid

Danger theory (DT)

DT [12]

O(N)

O(N)

It works way better than XED and SPRT as it exceeds false negative rates.

Getting started with learning in neural networks takes time.

Hybrid

Artificial intelligence-based

AI-DNN [60]

Corrective and perimeter defensive copy attacks based on hardware and perimeter defense systems (IDS/IPS) generating a high false positive rate can be reduced.

Expensive design and implementation costs.

Hybrid

Witness-based and multiple machine learning algorithms

Multiple machine learning models (MMLM) [61]

Ctot = Csn + Cw

Cm = Nn. L

Processing speed uses some machine learning algorithms and does not require additional memory.

Centralized

Context information sensed

Extended elliptic curve digital signature technique ECDSA [62]

O(N)

O($\sqrt{N}$)

Use three algorithms to location proof, Three algorithms work in this protocol, the site computation algorithm, the site proof creation algorithm, and the site verification algorithm, and this enhances the work of the protocol.

Distributed

Combine more than schemes

TDD/SDD [63]

-

-

There are no restrictions on the number and distribution of cloned nodes, and it incurs low computation and communication costs. TDD and SDD provide high detection accuracy and excellent resilience against cloned, collusive, smart nodes.

Distributed