Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT

Hadiseh Rezaei; Rahim Taheri; Ivan Jordanov; Stavros Shiaeles

doi:10.1109/CSR64739.2025.11130071

Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT

Hadiseh Rezaei, Rahim Taheri^*, Ivan Jordanov, Stavros Shiaeles

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

9 Downloads (Pure)

Abstract

Adopting Federated Learning (FL) in Intrusion Detection Systems (IDS) for IoT enhances privacy by decentralizing model training. However, FL is still vulnerable to adversarial threats, particularly Label-Flipping Attacks (LFA) that manipulate local training data to degrade model performance. This paper introduces the Adaptive Gradient-Guided Label Flipping Attack (Adapt-LFA), which strategically flips training labels using gradient-based optimization to maximize classification errors while minimizing detection risks. Evaluations on the CSE-CICIDS2018 and CICIoV2024 datasets show that Adapt-LFA reduces accuracy by 10% in Recurrent Neural Networks (RNN) and 13% in Convolutional Neural Networks (CNN), outperforming baseline LFA in disrupting FL-based IDS. These results highlight the effectiveness of the attack in degrading IDS performance within FL-based IoT environments.

Original language	English
Title of host publication	2025 IEEE International Conference on Cyber Security and Resilience (CSR)
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	407-412
Number of pages	6
ISBN (Electronic)	9798331535919
ISBN (Print)	9798331535926
DOIs	https://doi.org/10.1109/CSR64739.2025.11130071
Publication status	Published - 26 Aug 2025
Event	2025 IEEE International Conference on Cyber Security and Resilience - Chania, Crete, Greece Duration: 4 Aug 2025 → 6 Aug 2025 https://www.ieee-csr.org/#technical-program-committee

Conference

Conference	2025 IEEE International Conference on Cyber Security and Resilience
Country/Territory	Greece
City	Crete
Period	4/08/25 → 6/08/25
Internet address	https://www.ieee-csr.org/#technical-program-committee

Keywords

Federated Learning
Intrusion Detection Systems
Adversarial Attack
Label Flipping Attack

Access to Document

10.1109/CSR64739.2025.11130071

AGLFA_Post_PrintAccepted author manuscript (Post-print), 535 KBLicence: CC BY

Cite this

Rezaei, H., Taheri, R., Jordanov, I., & Shiaeles, S. (2025). Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT. In 2025 IEEE International Conference on Cyber Security and Resilience (CSR) (pp. 407-412). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CSR64739.2025.11130071

@inproceedings{cd2d44a35bed47f49672998ecdfcbd7a,

title = "Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT",

abstract = "Adopting Federated Learning (FL) in Intrusion Detection Systems (IDS) for IoT enhances privacy by decentralizing model training. However, FL is still vulnerable to adversarial threats, particularly Label-Flipping Attacks (LFA) that manipulate local training data to degrade model performance. This paper introduces the Adaptive Gradient-Guided Label Flipping Attack (Adapt-LFA), which strategically flips training labels using gradient-based optimization to maximize classification errors while minimizing detection risks. Evaluations on the CSE-CICIDS2018 and CICIoV2024 datasets show that Adapt-LFA reduces accuracy by 10\% in Recurrent Neural Networks (RNN) and 13\% in Convolutional Neural Networks (CNN), outperforming baseline LFA in disrupting FL-based IDS. These results highlight the effectiveness of the attack in degrading IDS performance within FL-based IoT environments.",

keywords = "Federated Learning, Intrusion Detection Systems, Adversarial Attack, Label Flipping Attack",

author = "Hadiseh Rezaei and Rahim Taheri and Ivan Jordanov and Stavros Shiaeles",

year = "2025",

month = aug,

day = "26",

doi = "10.1109/CSR64739.2025.11130071",

language = "English",

isbn = "9798331535926",

pages = "407--412",

booktitle = "2025 IEEE International Conference on Cyber Security and Resilience (CSR)",

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

address = "United States",

note = "2025 IEEE International Conference on Cyber Security and Resilience<br/> ; Conference date: 04-08-2025 Through 06-08-2025",

url = "https://www.ieee-csr.org/\#technical-program-committee",

}

Rezaei, H, Taheri, R , Jordanov, I & Shiaeles, S 2025, Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT. in 2025 IEEE International Conference on Cyber Security and Resilience (CSR). Institute of Electrical and Electronics Engineers Inc., pp. 407-412, 2025 IEEE International Conference on Cyber Security and Resilience
, Crete, Greece, 4/08/25. https://doi.org/10.1109/CSR64739.2025.11130071

Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT. / Rezaei, Hadiseh; Taheri, Rahim ; Jordanov, Ivan et al.
2025 IEEE International Conference on Cyber Security and Resilience (CSR). Institute of Electrical and Electronics Engineers Inc., 2025. p. 407-412.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT

AU - Rezaei, Hadiseh

AU - Taheri, Rahim

AU - Jordanov, Ivan

AU - Shiaeles, Stavros

PY - 2025/8/26

Y1 - 2025/8/26

N2 - Adopting Federated Learning (FL) in Intrusion Detection Systems (IDS) for IoT enhances privacy by decentralizing model training. However, FL is still vulnerable to adversarial threats, particularly Label-Flipping Attacks (LFA) that manipulate local training data to degrade model performance. This paper introduces the Adaptive Gradient-Guided Label Flipping Attack (Adapt-LFA), which strategically flips training labels using gradient-based optimization to maximize classification errors while minimizing detection risks. Evaluations on the CSE-CICIDS2018 and CICIoV2024 datasets show that Adapt-LFA reduces accuracy by 10% in Recurrent Neural Networks (RNN) and 13% in Convolutional Neural Networks (CNN), outperforming baseline LFA in disrupting FL-based IDS. These results highlight the effectiveness of the attack in degrading IDS performance within FL-based IoT environments.

AB - Adopting Federated Learning (FL) in Intrusion Detection Systems (IDS) for IoT enhances privacy by decentralizing model training. However, FL is still vulnerable to adversarial threats, particularly Label-Flipping Attacks (LFA) that manipulate local training data to degrade model performance. This paper introduces the Adaptive Gradient-Guided Label Flipping Attack (Adapt-LFA), which strategically flips training labels using gradient-based optimization to maximize classification errors while minimizing detection risks. Evaluations on the CSE-CICIDS2018 and CICIoV2024 datasets show that Adapt-LFA reduces accuracy by 10% in Recurrent Neural Networks (RNN) and 13% in Convolutional Neural Networks (CNN), outperforming baseline LFA in disrupting FL-based IDS. These results highlight the effectiveness of the attack in degrading IDS performance within FL-based IoT environments.

KW - Federated Learning

KW - Intrusion Detection Systems

KW - Adversarial Attack

KW - Label Flipping Attack

U2 - 10.1109/CSR64739.2025.11130071

DO - 10.1109/CSR64739.2025.11130071

M3 - Conference contribution

SN - 9798331535926

SP - 407

EP - 412

BT - 2025 IEEE International Conference on Cyber Security and Resilience (CSR)

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2025 IEEE International Conference on Cyber Security and Resilience<br/>

Y2 - 4 August 2025 through 6 August 2025

ER -

Adapt-LFA: adaptive gradient-guided label flipping attack against federated learning-based intrusion detection in IoT

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this