Student or educator? Sign up with your school email and get 6 months of Pro free.

The DOI matches its Crossref record; 26 of the 28 references it cites were found in a registry or an index; none of the 8 sources its claims cite is checked for support yet.

Comparison of ML/DL Approaches for Detecting DDoS Attacks in SDN

Open access

Abstract

Software-defined networking (SDN) presents novel security and privacy risks, including distributed denial-of-service (DDoS) attacks. In response to these threats, machine learning (ML) and deep learning (DL) have emerged as effective approaches for quickly identifying and mitigating anomalies. To this end, this research employs various classification methods, including support vector machines (SVMs), K-nearest neighbors (KNNs), decision trees (DTs), multiple layer perceptron (MLP), and convolutional neural networks (CNNs), and compares their performance. CNN exhibits the highest train accuracy at 97.808%, yet the lowest prediction accuracy at 90.08%. In contrast, SVM demonstrates the highest prediction accuracy of 95.5%. As such, an SVM-based DDoS detection model shows superior performance. This comparative analysis offers a valuable insight into the development of efficient and accurate techniques for detecting DDoS attacks in SDN environments with less complexity and time.

Overview

CoChat Summary

TL;DR

This paper compares SVM, KNN, DT, MLP, and CNN for detecting DDoS attacks in SDN using the CICIDS2017 and CICDDoS2019 datasets. SVM showed the best prediction accuracy and F1-score, while CNN had the highest training accuracy but lower prediction accuracy and much longer training/prediction times.

Key findings

  • SVM achieved the highest prediction accuracy, reported as 95.57% on CICDDoS2019 and 94.01% on CICIDS2017, with F1-score 0.97 and MCC 0.94; the abstract summarizes this as 95.5% prediction accuracy.
  • CNN had the highest training accuracy (97.81% on CICDDoS2019; 95.18% on CICIDS2017) but lower prediction accuracy (90.09% and 89.61%, respectively), and the longest training/prediction times (1800.59 s training, 157.22 s prediction).
  • DT performed strongly with 96.43% training accuracy and 95.31% prediction accuracy on CICDDoS2019, and a low training time of 0.45 s.
  • KNN had the lowest accuracy in both training and prediction (91.23%/90.61% on CICDDoS2019; 90.14%/89.16% on CICIDS2017), though its training time was very low at 0.033 s.

Methods

  • Five ML/DL classifiers—SVM, KNN, DT, MLP, and CNN—were trained and tested on CICIDS2017 and CICDDoS2019; CICIDS2017 contains 79 features and 6 attack methods, but only one day (Tuesday) with normal and attack traffic totaling 11 G was used, while CICDDoS2019 contains 84 features and 12 attack methods.
  • Preprocessing included linear interpolation for missing values in Flow Bytes/s and Flow Packets/s, feature selection that reduced the dataset to 50 features, and replacement of “Intf” values in FlowBytes/s and FlowPackets/s with zero for KNN, MLP, and SVM.
  • Performance was compared using train accuracy, prediction accuracy, F1-score, Matthews correlation coefficient (MCC), training time, and predict time; experiments used Jupyter Notebook 6.4.12 and Python 8.4.0 on a Windows 10 laptop with an Intel Core i5-4200U CPU, 12 GB RAM, and Intel HD Graphics 3000.

Limitations

  • The paper states that only one day’s worth of data from several days was analyzed, and that using the entire dataset with all six attack types is needed for a clear conclusion.
  • The paper states that the work is based on simulation and offline analysis, making it difficult to assess detection algorithms in real time and on actual networks.
  • The text makes evident inconsistencies in reported numbers across the abstract, discussion, conclusion, and tables—for example, CNN training/prediction accuracy appears as 97.808%/90.08% in the abstract but 97.81%/90.09% in the CICDDoS2019 table, and SVM prediction accuracy appears as 95.5% in the abstract versus 95.57% in the table.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/app13053033

Cites

26 of 26 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Process flow diagram.
Figure 1 Process flow diagram.CC-BY · From the PDF
Figure 2 Hyperplane SVM.
Figure 2 Hyperplane SVM.CC-BY · From the PDF
Figure 3 Multi-layer perceptron (MLP) diagram.
Figure 3 Multi-layer perceptron (MLP) diagram.CC-BY · From the PDF
Image 4
CC-BY · From the PDF
Image 5
CC-BY · From the PDF
Figure 7 Models' accuracy result.
Figure 7 Models' accuracy result.CC-BY · From the PDF
Figure 8 Models' time result.
Figure 8 Models' time result.CC-BY · From the PDF
Figure 9 Models' F1-score result.
Figure 9 Models' F1-score result.CC-BY · From the PDF
Figure 10 Models' MCC result.
Figure 10 Models' MCC result.CC-BY · From the PDF
Image 10
CC-BY · From the PDF
Figure 12 Models' F1-score comparison result.
Figure 12 Models' F1-score comparison result.CC-BY · From the PDF
Figure 13 Models' timing comparison result.
Figure 13 Models' timing comparison result.CC-BY · From the PDF

Verification

Claims the paper makes

0 Valid · 3 Not valid · 6 Not checked · 42 Own results and interpretations

23 more claims

Create a free account to see all 51 claims and to verify the 7 references still open.

See all claims — free account
  • It may state two separate claims as one. 15
  • Probably not a substantial claim. 10
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

5 of 7 · the rest aren’t loaded

2 more findings

Create a free account to see every finding of this check.

See all findings — free account

Citations

Cites

Show all 28 references

Research works printed but not linked: 3

  1. Roy, 2016, Random forest, support vector machine and nearest centroid methods for classifying network intrusion, Comput. Sci. Ser., 14, 9
  2. Imandoust, 2013, Application of k-nearest neighbor (knn) approach for predicting economic events: Theoretical background, Int. J. Eng. Res. Appl., 3, 605
  3. Balogun, 2017, An ensemble approach based on decision tree and bayesian network for intrusion detection, Comput. Sci. Ser., 15, 82

Cited by

See all 91

Metadata

Authors
Ali, T. E., Chong, Y.-W., & Manickam, S.
Affiliations
Universiti Sains Malaysia (MY)
Corresponding author
Chong, Yung-Wey
Venue
Applied Sciences
Publisher
MDPI AG
Published
2023-02-27
Volume
13
Issue
5
Pages
3033
ISSN
2076-3417 (ISSN-L)
OpenAlex
W4322502180
Cited by
91
References
28
Citations by year
2026: 18 · 2025: 30 · 2024: 35 · 2023: 8
FWCI
7.69
Type
journal article
Language
en
Open access
Locations
  • Applied Sciences (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Universiti Sains Malaysia (USM)
  • National Advanced IPv6 Center (NAv6)
Topics
  • Physical Sciences › Computer Science › Computer Networks and Communications › Network Security and Intrusion Detection
  • Physical Sciences › Computer Science › Computer Networks and Communications › Software-Defined Networks and 5G
  • Physical Sciences › Computer Science › Hardware and Architecture › Network Packet Processing and Optimization
Keywords
DDoS attack detection, software-defined networking, network security, support vector machine, convolutional neural networks, machine learning classification, deep learning, anomaly detection, model performance comparison

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-16, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-16 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Comparison of ML/DL Approaches for Detecting DDoS Attacks in SDN. https://app.cochat.ai/papers/10.3390/app13053033
The paper
Ali, T. E., Chong, Y.-W., & Manickam, S. (2023). Comparison of ML/DL Approaches for Detecting DDoS Attacks in SDN. Applied Sciences. https://doi.org/10.3390/app13053033

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.