Claim Missing Document
Check
Articles

Found 34 Documents
Search

Segmentasi Paru Tuberkulosis Pada Citra X-Ray Dada: Eksperimen Model Empat Arsitektur CNN Berbasis Deep Learning Hendra Hermansyah; Sawali Wahyu
KRESNA: Jurnal Riset dan Pengabdian Masyarakat Vol 6 No 1 (2026): Jurnal KRESNA Mei 2026
Publisher : DRPM Universitas Budi Luhur

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.36080/kresna.v6i1.319

Abstract

Tuberkulosis (TB) masih menjadi salah satu penyakit infeksi menular utama di dunia, dengan sekitar 10 juta kasus baru setiap tahun. Segmentasi paru secara otomatis dari citra chest X-ray (CXR) merupakan tahap prapemrosesan penting bagi sistem diagnosis berbantuan komputer. Penelitian ini menyajikan analisis komparatif empat arsitektur deep learning untuk segmentasi paru TB pada dataset CXR Shenzhen Hospital (n=662), dengan mengevaluasi lebih dari 30 metrik kinerja yang mencakup klasifikasi, kualitas segmentasi, dan metrik lanjutan. Empat arsitektur yang dibandingkan adalah U-Net klasik, U-Net+ResNet, RNGU-Net, dan MobileNetV2-UNet, menggunakan validasi silang 5-fold serta uji signifikansi statistik. Hasil penelitian menunjukkan bahwa RNGU-Net memberikan kinerja paling seimbang di antara keempat arsitektur, dengan akurasi 97,38%, recall 92,89%, dan F1-Score 94,59%, disertai tingkat false negative terendah (7,11%) dan kalibrasi kepercayaan terbaik (ECE=0,0385). U-Net menawarkan waktu pelatihan tercepat (4,20 menit) dengan pengorbanan akurasi yang minimal, sedangkan MobileNetV2-UNet menunjukkan penurunan recall yang signifikan (81,35%) sehingga membatasi kesesuaiannya untuk konteks klinis. Temuan ini memberikan rekomendasi berbasis bukti untuk penerapan segmentasi TB berbasis AI pada fasilitas layanan kesehatan dengan sumber daya terbatas.
Centralized Orchestration for Agent-Based Host Intrusion Detection System with Threat Intelligence Gede Ananda; Sawali Wahyu; Muhamad Hadi Arfian; Nugroho Budhi Santoso
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10390

Abstract

Cyber threats such as malware injection, web shell exploitation, phishing, and lateral movement increasingly challenge host-level security mechanisms. Conventional Host-Based Intrusion Detection Systems (HIDS) are commonly deployed in stand-alone configurations, resulting in limited cross-host visibility, delayed incident response, and lack of integration with external threat intelligence. This study proposes an agent-based HIDS architecture with centralized orchestration and real-time threat intelligence integration to improve detection accuracy and response efficiency. The system is developed using an experimental approach based on the NIST SP 800-61 Rev.2 incident handling framework, covering preparation, detection and analysis, containment and recovery, and post-incident evaluation. Each host deploys a lightweight agent that monitors file system activities, generates cryptographic hash values, and sends artifact metadata to a centralized orchestration server. The server performs parallel validation using external threat intelligence services and executes automated containment actions. Experimental results in a multi-host virtual environment show a False Positive Rate (FPR) of 3.2%, Mean Time to Detect (MTTD) of 4.8 seconds, and Mean Time to Respond (MTTR) of 6.5 seconds, with a 38% improvement compared to manual monitoring. These findings indicate that centralized orchestration combined with threat intelligence integration enhances detection precision, scalability, and incident response effectiveness in HIDS
Implementation of IT Security Operations Management Application for Cyber Security Threat Monitoring Saputra, Tengku Arya; Wahyu, Sawali; Arfian, Muhamad Hadi; Santoso, Nugroho Budi
JEPIN (Jurnal Edukasi dan Penelitian Informatika) Vol. 12 No. 1 (2026): Volume 12 No 1
Publisher : Program Studi Informatika

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26418/jp.v12i1.105673

Abstract

Peningkatan kompleksitas dan volume serangan siber mengharuskan organisasi untuk menerapkan mekanisme pemantauan keamanan terintegrasi dan real-time. Studi ini mengusulkan model pemantauan berbasis Operasi Keamanan (Security Operations) untuk meningkatkan efektivitas deteksi serangan siber dan respons insiden di lingkungan sektor keuangan. Sistem ini mengintegrasikan pemantauan firewall aplikasi web, pemantauan integritas file berbasis host, dan pengayaan intelijen ancaman siber dalam platform terpusat yang mampu melakukan tindakan respons otomatis, termasuk pemblokiran IP dan karantina file. Model ini dievaluasi menggunakan skenario serangan terkontrol yang meliputi serangan injeksi SQL, skrip lintas situs, eksekusi kode jarak jauh, dan unggahan file berbahaya. Hasil eksperimen menunjukkan bahwa semua serangan berhasil dideteksi dan diatasi. Sistem mencapai waktu deteksi berkisar antara 0 hingga 37 detik dan waktu respons antara 3 hingga 6 detik, dengan rata-rata waktu respons 4 detik. Pemantauan lintas lapisan dan penahanan otomatis mengurangi paparan serangan dan meningkatkan efisiensi penanganan insiden operasional. Temuan ini menunjukkan bahwa integrasi deteksi multi-lapisan dengan respons otomatis memberikan perbaikan yang dapat diukur dalam operasi keamanan dunia nyata. Kerangka kerja mini–Security Operations Center yang diusulkan menawarkan pendekatan pemantauan keamanan praktis bagi organisasi dengan sumber daya keamanan terbatas.
TELEGRAM-BASED CHATBOT DESIGN FOR ACADEMIC SERVICE AUTOMATION AT FASILKOM UEU USING A LOW-CODE PLATFORM Gabriela Marry Christiana Simanjuntak; Dewi Setiowati; Binastya Anggara Sekti; Sawali Wahyu
JIKO (Jurnal Informatika dan Komputer) Vol 9 No 2 (2026)
Publisher : Program Studi Teknik Informatika Universitas Khairun

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33387/jiko.v9i2.11579

Abstract

Administrative services at the Faculty of Computer Science, Universitas Esa Unggul, face persistent challenges due to high volumes of repetitive inquiries and limited operational hours, which may reduce service efficiency and responsiveness. This study aims to design, implement, and evaluate FasilkomAssist, a 24/7 Telegram-based chatbot developed using the ADDIE framework and a Low-Code/No-Code (LCNC) approach via the n8n workflow automation platform. The system integrates Google Gemini as a Large Language Model (LLM) with Google Sheets, Google Drive, Google Calendar, and Telegram to automate the dissemination of Thesis information and administrative procedures. Evaluation was conducted through Black-box testing, the System Usability Scale (SUS), and User Acceptance Testing (UAT) involving 30 respondents. Results indicate a 96% response accuracy across 17 valid testing scenarios, with an average response time of 4–6 seconds. The SUS score of 70.75 places the system within the “Acceptable” range (Grade C), indicating satisfactory usability with opportunities for optimization. Furthermore, UAT conducted with Faculty Administration Staff yielded a score of 84.67%, confirming functional adequacy and user acceptance. These findings demonstrate that an LCNC-based chatbot can improve efficiency, accessibility, and service quality in academic administrative environments without requiring advanced programming expertise.