Sarah Rosdiana Tambunan
Unknown Affiliation

Published : 4 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 4 Documents
Search

C Source code Obfuscation using Hash Function and Encryption Algorithm Sarah Rosdiana Tambunan; Nur Rokhman
IJCCS (Indonesian Journal of Computing and Cybernetics Systems) Vol 17, No 3 (2023): July
Publisher : IndoCEISS in colaboration with Universitas Gadjah Mada, Indonesia.

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.22146/ijccs.86118

Abstract

Obfuscation is a technique for transforming program code into a different form that is more difficult to understand. Several obfuscation methods are used to obfuscate source code, including dead code insertion, code transposition, and string encryption. In this research, the development of an obfuscator that can work on C language source code uses the code transposition method, namely randomizing the arrangement of lines of code with a hash function and then using the DES encryption algorithm to hide the parameters of the hash function so that it is increasingly difficult to find the original format. This obfuscator is specifically used to maintain the security of source code in C language from plagiarism and piracy. In order to evaluate this obfuscator, nine respondents who understand the C programming language were asked to deobfuscate the obfuscated source code manually. Then the percentage of correctness and the average time needed to perform the manual deobfuscation are observed. The evaluation results show that the obfuscator effectively maintains security and complicates the source code analysis.
Enhancing Credit Risk Classification Using LightGBM with Deep Feature Synthesis Tambunan, Sarah Rosdiana; Amalia, Junita; Sitorus, Kristina Margaret; Sibuea, Yehezchiel Abed Rafles; Hutabarat, Lucas Ronaldi
Journal of Information System and Informatics Vol 6 No 4 (2024): December
Publisher : Universitas Bina Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51519/journalisi.v6i4.902

Abstract

In the digital financial services era, Peer-to-Peer (P2P) lending has emerged as a significant innovation in fintech. However, credit risk remains a major concern due to the potential for payment failures, which can cause losses for platforms and investors. This research explores the impact of Deep Feature Synthesis (DFS) on credit risk classification and evaluates the performance of the Light Gradient Boosting Machine (LightGBM) algorithm with and without DFS. The data used in this study was sourced from Kaggle, a peer-to-peer lending company based in San Francisco, California, United States. The dataset contains 74 attributes, with a total of 887,379 rows. DFS automatically generates new attributes, while LightGBM is used for selecting the most important features, aiming to optimize credit risk predictions and simplify the model's complexity. The results of credit risk classification models using DFS and without it. Findings reveal that DFS enhances the accuracy of the credit risk classification, achieving a 0.99 accuracy rate compared to 0.97 without DFS, achieving a recall and F1-score of 0.94 and 0.96 with DFS and 0.68 and 0.81 without DFS. These results suggest that DFS is an effective feature engineering technique for boosting credit risk model performance. This research contributes significantly to the P2P lending industry by demonstrating that combining DFS with LightGBM can improve credit risk management, making it a valuable approach for financial platforms.
Enhancing Hate Speech Detection: Leveraging Emoji Preprocessing with BI-LSTM Model Amalia, Junita; Tambunan, Sarah Rosdiana; Purba, Susi Eva Maria; Simanjuntak, Walker Valentinus
Journal of Information System and Informatics Vol 7 No 2 (2025): June
Publisher : Universitas Bina Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51519/journalisi.v7i2.1147

Abstract

Microblogging platforms like Twitter enable users to rapidly share opinions, information, and viewpoints. However, the vast volume of daily user-generated content poses challenges in ensuring the platform remains safe and inclusive. One key concern is the prevalence of hate speech, which must be addressed to foster a respectful and open environment. This study explores the effectiveness of the Emoji Description Method (EMJ DESC), which enhances tweet classification by converting emojis into descriptive text or sentences. These descriptions are then encoded into numerical vector matrices that capture the meaning and emotional tone of each emoji. Integrated into a basic text classification model, these vectors help improve detection performance. The research examines how different emoji preprocessing strategies affect the performance of a BI-LSTM model for hate speech classification. Results show that removing emojis significantly reduces accuracy (68%) and weakens the model’s ability to distinguish between hate and non-hate speech, due to the loss of valuable semantic context. In contrast, retaining emoji semantics either through textual descriptions or embeddings boosts classification accuracy to 93% and 94%, respectively. The highest performance is achieved through emoji embedding, highlighting its ability to capture subtle non-verbal cues critically for accurate hate speech detection. Overall, the findings emphasize the importance of incorporating emoji-aware preprocessing techniques to enhance the effectiveness of social media content classification.
Indonesian automated short-answer grading using transformers-based semantic similarity Situmeang, Samuel; Tambunan, Sarah Rosdiana; Ginting, Lidia; Simamora, Wahyu Krisdangolyanti; ButarButar, Winda Sari
International Journal of Informatics and Communication Technology (IJ-ICT) Vol 14, No 3: December 2025
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijict.v14i3.pp1034-1043

Abstract

Automatic short answer grading (ASAG) systems offer a promising solution for improving the efficiency of reading literacy assessments. While promising, current Indonesian artificial intelligence (AI) grading systems still have room for improvement, especially when dealing with different domains. This study explores the effectiveness of large language models, specifically bidirectional encoder representations from transformers (BERT) variants, in conjunction with traditional hand-engineered features, to improve ASAG accuracy. We conducted experiments using various BERT models, hand-engineered features, text pre-processing techniques, and dimensionality reduction. Our findings show that BERT models consistently outperform traditional methods like term frequency-inverse document frequency (TF-IDF). IndoBERTLite-Base-P2 achieved the highest quadratic weighted kappa (QWK) score among the BERT variants. Integrating handengineered features with BERT resulted in a substantial enhancement of the QWK score. Utilizing comprehensive text pre-processing is a critical factor in achieving optimal performance. In addition, dimensionality reduction should be carefully used because it potentially removes semantic information.