Journal of Technology and Informatics (JoTI)

Simulated Phishing Attack and Forensic Analysis Using the D4I Framework: A Case Study on Kredivo

Muhammad Yusuf Halim — 2025-09-25

Phishing is a form of cyberattack where attackers deceive users into revealing sensitive information such as credentials or financial data, often through fake communication channels or websites. This threat is particularly critical in the financial technology (fintech) sector, where services rely heavily on digital transactions and user trust. This study presents a simulated phishing case targeting Kredivo users to evaluate the effectiveness of the Digital Forensics framework for Reviewing and Investigating cyber-attacks (D4I) in digital forensic analysis. The Cyber Kill Chain (CKC) model was employed to trace attacker behavior across seven phases, from weaponization to actions on objectives. Forensic data was acquired using MOBILedit Forensic Express from two smartphones, namely an iPhone 11 (iOS 15.8.1) and a Vivo Y21 (Android 8.1.0), which served as simulated evidence devices. Using the D4I framework, the investigation successfully identified and correlated key digital artifacts such as phishing links, OTP transmissions, and unauthorized access logs. These findings were organized into a visual chain of artifacts to reconstruct the full attack lifecycle. The results demonstrate that the D4I framework is effective in guiding structured forensic investigations and understanding attack patterns, supporting the enhancement of fintech security strategies.

Analisis Sentimen Pengguna X terhadap Perempuan di Lingkungan Kerja Menggunakan Algoritma Machine Learning

Muhammad Davit Hilal Fahri — 2025-09-25

Gender bias against women in the workplace persists, including within digital interactions on social media. This study analyzes user sentiment on Platform X regarding women in professional contexts using three machine learning algorithms: Naïve Bayes, Support Vector Machine (SVM), and Random Forest. A total of 2,336 tweets were collected using 14 gender-related keywords and labeled both automatically using the DistilBERT model and manually through contextual interpretation. The automatic dataset was imbalanced (1,823 negative, 479 positive), while the manual dataset was more balanced (1,196 negative, 1,106 positive). After preprocessing and TF-IDF feature extraction, the data were split using the train_test_split method. Evaluation metrics included accuracy, precision, recall, and F1-score. Random Forest achieved the highest accuracy (79%) on automatic labels but showed class imbalance (F1-score: 0.88 for negative, 0.08 for positive). Meanwhile, models trained on manual labels showed more balanced performance with accuracy between 57% and 59%. A web application prototype was developed using Flask to predict sentiment related to workplace gender issues. The findings highlight the importance of balanced labeling and appropriate algorithm selection to build fair and reliable sentiment analysis models, contributing to more inclusive digital discourse on gender equality.

Sistem Pengenalan Wajah Real-Time Menggunakan YOLOv7 untuk Akses Gedung TVRI Palembang Berbasis Web

Fatia Salsabilla Kyara — 2025-09-29

The development of information and communication technology today has had a significant impact on various aspects of life, including in the field of security. The use of face recognition is one of the facial recognition techniques, where the results of the camera capture will be matched with photos or facial curve textures that already exist in the database. The system is widely applied using various methods and artificial intelligence, one of which is YOLO (You Only Look Once). The purpose of this study is to design, develop, and identify challenges in implementing a real-time facial recognition system using web-based YOLOv7 that can detect the faces of people entering the TVRI Palembang building, then photos and times when a person's face is not detected will be stored in the database. The data used comes from literature studies, data collection obtained from photos of TVRI television station employees' faces, software design with technology selection, user interface design, and algorithm structures that will be used. After going through these stages, a system implementation was carried out for the application of the system and analysis of the data results obtained. The results showed that the face detection system using YOLOv7 showed very good performance. In 100 training epochs, the system achieved 96,6% face detection accuracy and 90% face recognition accuracy, successfully identifying almost all registered faces and detecting faces in real time. This system produces high accuracy in detecting faces and almost all faces that should be recognized are successfully detected.

Implementasi Sistem Keamanan Brankas Berbasis Face Recognition Menggunakan Algoritma YOLO dengan Verifikasi Fingerprint

Amanda Tsabita Putri Tata — 2025-09-29

Current technological developments make security crucial in protecting important documents. Safes are commonly used for storage but remain vulnerable to burglary despite conventional locks. Based on data from the Central Bureau of Statistics (BPS) in 2023, aggravated theft was the most frequent crime with 62,872 reported cases. This highlights the need to improve safe security by adding biometric techniques such as face recognition and fingerprint verification. This study proposes a layered security system combining face recognition using the YOLO algorithm and fingerprint sensors. The system uses an ESP32-CAM to capture facial images and an ESP32 microcontroller to control a solenoid lock, fingerprint sensor and buzzer alarm. Face recognition testing on two users showed, the trained YOLO model achieved an accuracy of 83.33%, precision of 83,33% and recall of 100%. from 12 trials, with two failures due to poor lighting conditions. Fingerprint testing on 10 samples, five fingers from each of two users, showed successful recognition of all fingerprints with an average response time of 1.41 seconds. The integration of face and fingerprint biometrics significantly enhances safe security and minimizes unauthorized access risks.

Prediksi Stunting pada Anak Balita Menggunakan Algoritma Extreme Gradient Boosting dan Bayesian Optimization

Rangga Yoga Pratama — 2025-10-31

Stunting is a chronic malnutrition condition affecting children under five years that impairs cognitive development, physical growth, and future productivity. This study develops a stunting risk prediction model using the Extreme Gradient Boosting (XGBoost) algorithm with hyperparameter tuning and data balancing techniques. The dataset from Kaggle contains 120,998 records with variables including age, gender, height, and nutritional status. The methodology encompasses data preprocessing for outlier handling, categorical encoding, and feature extraction based on height thresholds. Feature selection utilized ANOVA F-test, while Exploratory Data Analysis identified height as the most influential attribute. To address class imbalance, Synthetic Minority Over-sampling Technique (SMOTE) was implemented, followed by Bayesian Optimization for hyperparameter tuning. Model evaluation was conducted using various data splits (80:20, 70:30, 60:40, 50:50) with metrics including accuracy, precision, recall, and F1-score. Results demonstrate that the optimized XGBoost model achieved exceptional performance with 0,982% accuracy, 0,973% precision, 0.979% recall, and 0,976% F1-score, consistently across all data configurations. The combination of XGBoost with Bayesian Optimization and SMOTE proves highly effective in handling imbalanced classification tasks. These findings highlight machine learning's potential in supporting public health initiatives through accurate early identification and targeted intervention for stunting prevention.

Performance Analysis of Provider and Riverpod State Management Library on Flutter Applications

Jonathan Aditya Puryanto — 2025-10-31

State management libraries are essential components in Flutter app development. This research aims to compare the performance of the state management library Provider and its successor, Riverpod, to assist Flutter developers in choosing the right solution. Two versions of the MovieDB app were built, each utilizing Provider and Riverpod. Performance testing was conducted using three metrics: CPU Utilization, Memory Usage, and Execution Time, across three data volumes (1,000, 5,000, and 10,000). The results showed that CPU Utilization varied by only 0.1–0.2% with Riverpod being slightly more efficient at 1,000 and 10,000 data volumes. Execution Times also showed minimal differences, with Riverpod being marginally faster by approximately 0.01 seconds at 5,000 and 10,000 data volumes. Riverpod excelled in Memory Usage, demonstrating an average reduction of about 3–6% across all data volumes, particularly at higher data volumes. In conclusion, the performance of both libraries is fundamentally similar, but Riverpod is offers better memory efficiency and architectural flexibility. Therefore, Riverpod is recommended for new projects, while Provider remains a viable option for stable existing applications that already use it.

Sentiment Perspective of Government's Free Nutritious Meal Policy on Social Media X using Indo-BERT and Bi-LTSM

Pungkas Subarkah — 2025-10-31

This research has the potential to make an important contribution to the development of computationally-based sentiment analysis, especially in the context of government policies regarding the Free Meal Program that will be implemented throughout Indonesia. This research was conducted using Indo-BERT and Bi-LSTM algorithms. These approaches were used to categorize emotions into three groups: neutral, negative, and positive. Data is obtained from posts on social media X, then after processing the data, it will be applied to both algorithms, namely Indo-BERT and Bi-LSTM. The research findings show that the model's performance in determining the public sentiment of government policies. Validation and valuation were conducted using the f1 score, recall, and precision metrics. The evaluation findings show that the Indo-BERT algorithm is better than the Bi-LSTM algorithm with an accuracy value of 80% for Indo-BERT and 78% for the accuracy value of the Bi-LSTM algorithm, and the Indo-BERT accuracy value is included in the good classification accuracy value. The sentiment analysis results are also represented by word clouds for each positive, negative and neutral class, providing an intuitive picture of the words frequently used in public discourse on free nutritious meals.

Segmentasi dan Klasifikasi Risiko Perdarahan dengan Algoritma K-Means dan Naive Bayes Berdasarkan Data Klinis dan Transfusi

Aryanti Aryanti — 2025-10-31

Bleeding risk assessment is essential in clinical decision-making, especially for patients undergoing frequent blood transfusions. This study presents a machine learning approach combining K-Means clustering and Gaussian Naive Bayes classification to assess bleeding risk based on clinical and transfusion history data. Patients were categorised into K-Means clusters, with the ideal number of clusters established by the Elbow Method and Silhouette Score. PCA visualisation demonstrated distinct distinctions among clusters. Cluster 0 contained patients with higher transfusion volume and frequency, showing significantly higher bleeding risk. Subsequently, the Naive Bayes classifier was trained on clinical features to predict bleeding risk and categorized into two risk levels. The model achieved 85.45 percent accuracy on training data and 86.67 percent on testing data, with the highest predictive accuracy observed in Cluster 0 (95.65 percent). These results highlight the potential of combining unsupervised and supervised learning techniques to enhance bleeding risk stratification and support better transfusion management.

Customer Satisfaction, Social Influence, and Facilitating Conditions Affecting Use Behavior in Ride-Hailing

Erwin Sutomo — 2025-10-31

The rapid expansion of online ride-hailing services has transformed urban mobility; however, sustaining user engagement remains a major challenge due to the interplay of psychological and technical factors. Despite widespread adoption, the determinants influencing users’ continued use of such applications remain inconsistent and not fully understood. Therefore, this study examines the effects of Customer Satisfaction (CS), Social Influence (SI), and Facilitating Conditions (FC) on Use Behavior (UB) of online transportation applications, with Behavioral Intention (BI) as a mediating variable. The research was conducted in Surabaya with 150 active users of online ride-hailing services, and data were analyzed using Structural Equation Modeling–Partial Least Squares (SEM-PLS). The findings indicate that FC significantly affects both BI and UB, while CS and SI have significant positive effects on BI but not directly on UB. Moreover, BI is identified as the strongest predictor of UB and acts as a significant mediator in the relationships CS → BI → UB, SI → BI → UB, and FC → BI → UB. These results highlight the central role of intention in linking psychological and technical factors to actual use behavior. The study suggests that service providers enhance customer satisfaction, utilize social influence, and strengthen facilitating conditions to sustain user engagement.