國立虎尾科技大學 |

Privacy-Preserving Network Anomaly Detection via Homomorphic Encryption /

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Privacy-Preserving Network Anomaly Detection via Homomorphic Encryption // Yeqi Shi.
作者:	Shi, Yeqi,
面頁冊數:	1 electronic resource (86 pages)
附註:	Source: Masters Abstracts International, Volume: 85-05.
Contained By:	Masters Abstracts International85-05.
標題:	Electrical engineering. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30688456
ISBN:	9798380837637

Privacy-Preserving Network Anomaly Detection via Homomorphic Encryption /
Shi, Yeqi,

Privacy-Preserving Network Anomaly Detection via Homomorphic Encryption /Yeqi Shi. - 1 electronic resource (86 pages)

Source: Masters Abstracts International, Volume: 85-05.

Network Anomaly Detection refers to the detection of abnormal network activities due to security breaches or other network issues. Traditional methods of detection often rely on known signatures of anomalies. More recently, the advancement of machine learning has become a more robust mechanism to address this issue. However, the nature of network anomaly detection directly contradicts with the notion of data privacy and confidentiality. Being able to detect anomalies while preserving the privacy of the data has become a hotly researched topic.This thesis proposes a privacy-preserving Random Forest classifier tailored for anomaly detection settings. The system leverages Fully Homomorphic Encryption (FHE) encryption to achieve privacy-preserving classification on encrypted data. This work improves the computation and communications costs compared to previous works. Furthermore, the system is hardware-accelerated by GPUs, resulting in a speedup of up to 160x over a single-threaded CPU. Importantly, the optimization techniques place special emphasis on the scalability of the overall system, so that it can process higher throughput by horizontally scaling the hardware resources. This thesis also proposes a client-server-based protocol for generic random forest predictions, suitable for use cases such as Machine-Learning-as-a-Service (MLaaS).The privacy-preserving Random Forest was evaluated by common classification metrics on a network detection dataset, CIC-DS-2017. Experiments have shown that our system is able to sustain a detection throughput of 15 Kbps with high detection accuracy (F1 > 0.90) on a system with an Nvidia A100 GPU.

English

ISBN: 9798380837637Subjects--Topical Terms:

596380
Electrical engineering.
Subjects--Index Terms:

Fully Homomorphic Encryption

Privacy-Preserving Network Anomaly Detection via Homomorphic Encryption /
LDR:03022nam a22003973i 4500 001 1157735
005 20250603111404.5
006 m o d
007 cr|nu||||||||
008 250804s2023 miu||||||m |||||||eng d
020 $a 9798380837637
035 $a (MiAaPQD)AAI30688456
035 $a AAI30688456
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Shi, Yeqi, $e author. $0 (orcid)0009-0004-5184-9589 $3 1483995
245 1 0 $a Privacy-Preserving Network Anomaly Detection via Homomorphic Encryption / $c Yeqi Shi.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2023
300 $a 1 electronic resource (86 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Masters Abstracts International, Volume: 85-05.
500 $a Advisors: Gulak, Glenn Committee members: Li, Baochun; Hatzinakos, Dimitrios.
502 $b M.A.S. $c University of Toronto (Canada) $d 2023.
520 $a Network Anomaly Detection refers to the detection of abnormal network activities due to security breaches or other network issues. Traditional methods of detection often rely on known signatures of anomalies. More recently, the advancement of machine learning has become a more robust mechanism to address this issue. However, the nature of network anomaly detection directly contradicts with the notion of data privacy and confidentiality. Being able to detect anomalies while preserving the privacy of the data has become a hotly researched topic.This thesis proposes a privacy-preserving Random Forest classifier tailored for anomaly detection settings. The system leverages Fully Homomorphic Encryption (FHE) encryption to achieve privacy-preserving classification on encrypted data. This work improves the computation and communications costs compared to previous works. Furthermore, the system is hardware-accelerated by GPUs, resulting in a speedup of up to 160x over a single-threaded CPU. Importantly, the optimization techniques place special emphasis on the scalability of the overall system, so that it can process higher throughput by horizontally scaling the hardware resources. This thesis also proposes a client-server-based protocol for generic random forest predictions, suitable for use cases such as Machine-Learning-as-a-Service (MLaaS).The privacy-preserving Random Forest was evaluated by common classification metrics on a network detection dataset, CIC-DS-2017. Experiments have shown that our system is able to sustain a detection throughput of 15 Kbps with high detection accuracy (F1 > 0.90) on a system with an Nvidia A100 GPU.
546 $a English
590 $a School code: 0779
650 4 $a Electrical engineering. $3 596380
650 4 $a Computer engineering. $3 569006
653 $a Fully Homomorphic Encryption
653 $a Network anomaly detection
653 $a Machine learning
690 $a 0464
690 $a 0544
710 2 $a University of Toronto (Canada). $b Electrical and Computer Engineering. $3 1148628
720 1 $a Gulak, Glenn $e degree supervisor.
773 0 $t Masters Abstracts International $g 85-05.
790 $a 0779
791 $a M.A.S.
792 $a 2023
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30688456