Text

Komparasi model shallow machine learning dan deep learning untuk mendeteksi URL Phishing

---

Jumlah serangan phishing terus meningkat sejak wabah COVID-19 pada akhir tahun 2019. Phishing merupakan salah satu cara untuk mencuri informasi kredensial seseorang. Dalam laporan tren aktivitas phishing yang dikeluarkan oleh Anti-Phishing Working Group (APWG), kasus phishing global terus meningkat sepanjang tahun 2021 hingga kuartal pertama 2022. Dalam mengurangi dan menanggulangi pengaruh negatif dari URL (Uniform Resource Locator) phishing, maka pemerintah dalam hal ini Badan Siber dan Sandi Negara (BSSN) melalui Pusat Pengkajian dan Pengembangan Teknologi Keamanan Siber dan Sandi (Puskajibang Tekkamsisan) melakukan kajian
terhadap deteksi URL phishing. Metode yang digunakan adalah algoritma shallow machine learning seperti Logistic Regression dan Multinomial Naive Bayes. Penelitian ini dilakukan untuk membandingkan algoritma shallow machine learning dan deep learning dalam melakukan klasifikasi URL phishing. Penelitian dimulai dengan melakukan feature extraction dari dataset yang digunakan. Tahap selanjutnya yaitu pemodelan yang untuk algoritma deep learning. Model deep learning yang telah dibuat diuji dan dibandingkan dengan model shallow machine learning. Hasil dari penelitian ini menunjukkan bahwa algoritma deep learning kombinasi LSTM dan GRU lebih unggul dari Random Forest, dengan hasil akhir akurasi deep learning 98,1% sedangkan Random Forest 97.4%, sehingga algoritma yang digunakan untuk tahap selanjutnya adalah deep learning. Tahap akhir dari penelitian ini yaitu implementasi model prediksi dalam bentuk website menggunakan framework Flask dengan hasil
klasifikasi berupa skor probabilitas URL terdeteksi phishing. --

The number of phishing attacks has continued to increase since the COVID-19 at the end of 2019. Phishing is one way to steal someone's credential information. In the report on trends in phishing activity released by the AntiPhishing Working Group (APWG), global phishing cases continued to increase throughout 2021 until the first quarter of 2022. In reducing and overcoming the negative influence of URL (Uniform Resource Locator) phishing, the government in this case The National Cyber and Password Agency (BSSN) through the Center for the Assessment and Development of Cyber and Password Security Technology (Puskajibang Tekkamsisan) conducted a study on the detection of phishing URLs. The method used is traditional machine learning algorithms such as Logistic Regression and Multinomial Naive Bayes. This study was conducted to compare traditional machine learning and deep learning algorithms in classifying phishing URLs. The study began by performing feature extraction from the dataset used. The next stage is modeling for deep learning algorithms. The deep learning models that have been created are tested and compared to traditional machine learning models. The results of this study indicate that the deep learning algorithm is superior to Random Forest, with the final result of deep learning accuracy being 98.1% while Random Forest is 97.4%, so the algorithm used for the next stage is deep learning. The final stage of this research is the implementation of a prediction model in the form of a website using the Flask framework with the classification results in the form of a URL probability score being detected by phishing.

---

Availability

No copy data

Detail Information

Series Title

-

Call Number

2022 NIZ k

Publisher

Bogor : Poltek SSN., 2022

Collation

xvii, 47hlm.

Language

Indonesia

ISBN/ISSN

--

Classification

--

Content Type

-

Media Type

-

Carrier Type

-

Edition

--

Subject(s)

Phishing
deep learning
Machine learning

Specific Detail Info

-

Statement of Responsibility

Nizam Aditya Zuhayr

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment