Hasil Pencarian  ::  Simpan CSV :: Kembali

"Serangan jaringan semakin beragam seiring berkembangnya internet. Dalam menghadapiserangan-serangan tersebut, diperlukan juga pengembangan sistem keamanan internetterhadap pengguna salah satunya adalah IDS. Intrusion detection system (IDS) merupakansistem keamanan dalam mengawasi aktivitas jaringan yang berbahaya bagi pengguna.Metode yang umum digunakan yaitu signature-based IDS. Signature-based IDSmenggunakan daftar serangan siber yang diketahui dalam menentukan jaringan berbahayaatau normal. Akan tetapi, IDS hanya mengetahui serangan yang diketahui saja danmembutuhkan input secara manual untuk mengubah daftar serangan sehingga tidak efektifdalam mengatasi serangan yang tidak ketahui. Oleh karena itu, penelitian ini berfokus padapengembangan IDS dengan pendekatan machine learning menggunakan model autoencoderuntuk reduksi dimensi dan pengaruhnya terhadap model IDS. Autoencoder yang digunakanpada penelitian ini terdapat 2 model yaitu non-symmetric deep autoencoder (NDAE) danmodifikasi dari NDAE menggunakan metode variational autoencoder (VAE) yang disebutsebagai V-NDAE, serta model PCA. Modifikasi NDAE bertujuan untuk mengambilinformasi penting dengan menggunakan distribusi probabilistik sehingga menjadi data yangberkualitas untuk pelatihan model IDS. Pengujian reduksi dimensi dari model-model inidilakukan dengan melatih model IDS yaitu model random forest. Penelitian ini dilakukanpada 2 dataset yang berbeda yaitu dataset CICIDS2017 dan dataset dari simulasi seranganjaringan. Metrik yang digunakan adalah metrik accuracy, precision, recall, F-1 score, ROCcurve. Berdasarkan pengujian yang telah dilakukan terhadap dataset CICIDS2017, modelNDAE memiliki nilai rata-rata akurasi validasi sebesar 90.85% sehingga memiliki nilai yanglebih besar daripada model V-NDAE yang memiliki nilai rata-rata akurasi validasi sebesar87.65%. Pelatihan model NDAE menggunakan hyperparameter yang paling optimal yaitudengan optimizer RMSProp dan batch size sebesar 128. Pada pengujian terhadap datasetdari simulasi serangan jaringan, model NDAE memiliki performa yang lebih baik daripadamodel V-NDAE dan model PCA. Model NDAE memiliki nilai rata-rata akurasi validasisebesar 94.66% dan model V-NDAE memiliki nilai rata-rata akurasi validasi sebesar66.32%. Pelatihan model NDAE menggunakan hyperparameter yang paling optimal yaitudengan optimizer Adam dan batch size sebesar 32.

---

The variety of network attacks increases as the internet evolves. In dealing with these attacks,

the development of an internet security system for users is necessary, one of which is IDS.

An intrusion detection system (IDS) is a security system designed to monitor network

activity that is dangerous for users. The commonly used method is signature-based IDS.

Signature-based IDS uses a signature database of known cyber attacks to determine whether

a network is dangerous or normal. However, this IDS only recognizes known attacks and

requires manual input to change the signature database of attacks, making it ineffective in

dealing with unknown attacks. Therefore, this research focuses on developing an IDS using

a machine learning approach, specifically using an autoencoder model for dimensionality

reduction and its impact on the IDS model. The models used in this research consists of a

non-symmetric deep autoencoder (NDAE), modification of NDAE using the variational

autoencoder (VAE) method, and PCA model. The modified NDAE can capture important

information from the latent distribution, which helps stabilize the training of the model.

Dimensionality reduction testing for both models is performed by training an IDS model,

specifically a random forest model. This research is conducted on two different datasets: the

CICIDS2017 dataset and a dataset from network attack simulations. The evaluation metrics

used are accuracy, precision, recall, F-1 score, and ROC curve. Based on the testing

performed on the CICIDS2017 dataset, the NDAE model achieves an average validation

accuracy of 90.85%, which is higher than the average validation accuracy of 87.65% for the

V-NDAE model and PCA model. The NDAE model's training is done using the most optimal

hyperparameters, specifically the RMSProp optimizer and a batch size of 128. In the testing

on the dataset from network attack simulations, the NDAE model outperforms the V-NDAE

model and PCA model. The NDAE model achieves an average validation accuracy of

94.66%, while the V-NDAE model achieves an average validation accuracy of 66.32%. The

NDAE model's training is done using the most optimal hyperparameters, specifically the

Adam optimizer and a batch size of 32."

"This book highlights research in linking and mining data from across varied data sources. The authors focus on recent advances in this burgeoning field of multi-source data fusion, with an emphasis on exploratory and unsupervised data analysis, an area of increasing significance with the pace of growth of data vastly outpacing any chance of labeling them manually. The book looks at the underlying algorithms and technologies that facilitate the area within big data analytics, it covers their applications across domains such as smarter transportation, social media, fake news detection and enterprise search among others. This book enables readers to understand a spectrum of advances in this emerging area, and it will hopefully empower them to leverage and develop methods in multi-source data fusion and analytics with applications to a variety of scenarios.Includes advances on unsupervised, semi-supervised and supervised approaches to heterogeneous data linkage and fusion; Covers use cases of analytics over multi-view and heterogeneous data from across a variety of domains such as fake news, smarter transportation and social media, among others;Provides a high-level overview of advances in this emerging field and empowers the reader to explore novel applications and methodologies that would enrich the field."

"The objective of this book is to provide the reader with a comprehensive survey of the topic compressed sensing in information retrieval and signal detection with privacy preserving functionality without compromising the performance of the embedding in terms of accuracy or computational efficiency. The reader is guided in exploring the topic by first establishing a shared knowledge about compressed sensing and how it is used nowadays. Then, clear models and definitions for its use as a cryptosystem and a privacy-preserving embedding are laid down, before tackling state-of-the-art results for both applications. The reader will conclude the book having learned that the current results in terms of security of compressed techniques allow it to be a very promising solution to many practical problems of interest. The book caters to a broad audience among researchers, scientists, or engineers with very diverse backgrounds, having interests in security, cryptography and privacy in information retrieval systems. Accompanying software is made available on the authors’ website to reproduce the experiments and techniques presented in the book. The only background required to the reader is a good knowledge of linear algebra, probability and information theory."