Intrusion Prediction using LSTM and GRU with UNSW-NB15

Seongsoo Kim; Lei Chen; Jongyeop Kim

doi:10.1109/ComComAp53641.2021.9652926

Intrusion Prediction using LSTM and GRU with UNSW-NB15

Seongsoo Kim, Lei Chen, Jongyeop Kim

Information Technology

Georgia Southern University

Research output: Contribution to book or proceeding › Conference article › peer-review

5 Scopus citations

Abstract

This study proposes a deep learning model for predicting the timing of attacks by applying Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) to an intrusion detection system (IDS) dataset UNSW-NB15. We applied the finite state machine concept to convert floating-point values to equivalent binary ones to increase model accuracy. As a result, the accuracy of the GRU and LSTM by average 13% and 18% respectively in terms of weighted F1 score.

Original language	English
Title of host publication	2021 Computing, Communications and IoT Applications, ComComAp 2021
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	101-106
Number of pages	6
ISBN (Electronic)	9781665427975
DOIs	https://doi.org/10.1109/ComComAp53641.2021.9652926
State	Published - 2021
Event	2021 Computing, Communications and IoT Applications, ComComAp 2021 - Shenzhen, China Duration: Nov 26 2021 → Nov 28 2021

Publication series

Name	2021 Computing, Communications and IoT Applications, ComComAp 2021

Conference

Conference	2021 Computing, Communications and IoT Applications, ComComAp 2021
Country/Territory	China
City	Shenzhen
Period	11/26/21 → 11/28/21

Keywords

Big Data
Deep Learning
GRU
LSTM
UNSW-NB15

Access to Document

10.1109/ComComAp53641.2021.9652926

Cite this

@inproceedings{42a78624cefa475883abd7fc026c7775,

title = "Intrusion Prediction using LSTM and GRU with UNSW-NB15",

abstract = "This study proposes a deep learning model for predicting the timing of attacks by applying Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) to an intrusion detection system (IDS) dataset UNSW-NB15. We applied the finite state machine concept to convert floating-point values to equivalent binary ones to increase model accuracy. As a result, the accuracy of the GRU and LSTM by average 13% and 18% respectively in terms of weighted F1 score.",

keywords = "Big Data, Deep Learning, GRU, LSTM, UNSW-NB15",

author = "Seongsoo Kim and Lei Chen and Jongyeop Kim",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 2021 Computing, Communications and IoT Applications, ComComAp 2021 ; Conference date: 26-11-2021 Through 28-11-2021",

year = "2021",

doi = "10.1109/ComComAp53641.2021.9652926",

language = "English",

series = "2021 Computing, Communications and IoT Applications, ComComAp 2021",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "101--106",

booktitle = "2021 Computing, Communications and IoT Applications, ComComAp 2021",

address = "United States",

}

Kim, S, Chen, L & Kim, J 2021, Intrusion Prediction using LSTM and GRU with UNSW-NB15. in 2021 Computing, Communications and IoT Applications, ComComAp 2021. 2021 Computing, Communications and IoT Applications, ComComAp 2021, Institute of Electrical and Electronics Engineers Inc., pp. 101-106, 2021 Computing, Communications and IoT Applications, ComComAp 2021, Shenzhen, China, 11/26/21. https://doi.org/10.1109/ComComAp53641.2021.9652926

Intrusion Prediction using LSTM and GRU with UNSW-NB15. / Kim, Seongsoo; Chen, Lei ; Kim, Jongyeop.
2021 Computing, Communications and IoT Applications, ComComAp 2021. Institute of Electrical and Electronics Engineers Inc., 2021. p. 101-106 (2021 Computing, Communications and IoT Applications, ComComAp 2021).

Research output: Contribution to book or proceeding › Conference article › peer-review

TY - GEN

T1 - Intrusion Prediction using LSTM and GRU with UNSW-NB15

AU - Kim, Seongsoo

AU - Chen, Lei

AU - Kim, Jongyeop

PY - 2021

Y1 - 2021

N2 - This study proposes a deep learning model for predicting the timing of attacks by applying Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) to an intrusion detection system (IDS) dataset UNSW-NB15. We applied the finite state machine concept to convert floating-point values to equivalent binary ones to increase model accuracy. As a result, the accuracy of the GRU and LSTM by average 13% and 18% respectively in terms of weighted F1 score.

AB - This study proposes a deep learning model for predicting the timing of attacks by applying Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) to an intrusion detection system (IDS) dataset UNSW-NB15. We applied the finite state machine concept to convert floating-point values to equivalent binary ones to increase model accuracy. As a result, the accuracy of the GRU and LSTM by average 13% and 18% respectively in terms of weighted F1 score.

KW - Big Data

KW - Deep Learning

KW - GRU

KW - LSTM

KW - UNSW-NB15

UR - http://www.scopus.com/inward/record.url?scp=85124631072&partnerID=8YFLogxK

U2 - 10.1109/ComComAp53641.2021.9652926

DO - 10.1109/ComComAp53641.2021.9652926

M3 - Conference article

AN - SCOPUS:85124631072

T3 - 2021 Computing, Communications and IoT Applications, ComComAp 2021

SP - 101

EP - 106

BT - 2021 Computing, Communications and IoT Applications, ComComAp 2021

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2021 Computing, Communications and IoT Applications, ComComAp 2021

Y2 - 26 November 2021 through 28 November 2021

ER -

Intrusion Prediction using LSTM and GRU with UNSW-NB15

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this