Electronic Theses and Dissertation

Dalam jaringan masih mempunyai berbagai permasalahan di keamanan dikarenakan jaringan rentan terhadap agresi. Satu dari beberapa agresi yang terjadi adalah Distributed Denial of Service (DDoS) yang merupakan serangan yang menyerang bandwidth secara penuh sehingga menyebabkan jaringan down sehingga sebuah host dan service tidak dapat diakses. Beberapa tahun terakhir ini, deep learning berkembang sangat pesat. Teknologi ini sudah diterapkan pada berbagai perangkat lunak AI seperti pengenalan gambar, pengambilan gambar, mesin pencari dan pencarian info, dan pemrosesan bahasa alami. Penelitian ini menerapkan Deep Learning dengan arsitektur (Convolutional Neural Network) CNN VGG19 karena dianggap pantas untuk diimplementasikan karena karakteristiknya yang mampu mempelajari fitur yang lebih tidak berbentuk, mengurangi kompleksitas pelatihan model, akurasi yang menjanjikan, dan kemampuan untuk menangani deretan data yang besar. Setiap aktivitas penerimaan dan pengiriman packet akan tersimpan dan terekam dalam wireshark. Hasil log tersebut akan diolah, dan diambil beberapa nilai parameter seperti source IP, destination IP, protocol, packet length, dll sebelum akhirnya dapat menjadi data input untuk CNN VGG-19. Dari penilitian ini, didapat keluaran berupa sistem pendeteksian serangan DDoS dengan menggunakan penerapan metode Deep Learning. Berdasarkan penelitian yang telah dilakukan, pada pengujian dengan variasi data dengan jumlah 5999, 13000, dan 17000 data didapat rata-rata akurasi sebesar 99,73% dan rata-rata loss sebesar 2,20%. Dan pada pengujian dengan variasi iterasi, dengan jumlah iterasi 10, 20, dan 30 dan dengan data sebanyak 5000, 10000, dan 15000 data, maka didapat rata-rata akurasi sebesar 99,63% dan rata-rata loss sebesar 4,03%.

The network still has various problems in security because the network is vulnerable to attacks. One of the several attacks that occur is Distributed Denial of Service (DDoS) which is an attack that attacks the full bandwidth, causing the network to down so that a host and service cannot be accessed. In recent years, deep learning has developed very rapidly. Deep learning is part of Machine Learning which consists of many layers (hidden layer). This technology is already being applied to various AI software such as image recognition, image capture, search engines and info retrieval, and natural language processing. This study applies Deep Learning with the CNN VGG19 architecture because it is considered appropriate to be implemented because of its characteristics of being able to learn more amorphous features, reducing the complexity of model training, promising accuracy, and the ability to handle large data sets. Every activity of receiving and sending packets will be stored and recorded in wireshark. The log results will be processed, and several parameter values are taken such as source IP, destination IP, protocol, packet length, etc. before finally being able to become input data for CNN VGG-19. From this research, the output is in the form of a DDoS attack detection system using the application of the Deep Learning method. Based on the research that has been done, in testing with data variations with the amount of 5999, 13000, and 17000 data obtained an average accuracy of 99.73% and an average loss of 2.20%. And in the test with a variety of iterations, with the number of iterations 10, 20, and 30 and with data as much as 5000, 10000, and 15000 data, the average accuracy is 99.63% and the average loss is 4.03%.