Malicious and Benign URL Dataset Generation Using Character-Level LSTM Models
Bibliographic record
Abstract
As technologies advance, so do the attacks on them. Cybersecurity plays a significant role in society to protect everyone. Malicious URLs are links designed to promote scams, attacks, and frauds. Companies often have web filtering algorithms that will blacklist specific URLs as malicious; however, due to privacy concerns, they will not give outside entities access to their cybersecurity data. Unfortunately, this lack of data creates a dire need for more data in cybersecurity research and machine learning applications. This paper proposes using machine learning to generate new synthetic URLs characteristically indistinguishable from the data they replace. To do this two character-level long short-term memory (LSTM) models were trained, one to generate malicious URLs and one to generate benign URLs. To assess the quality of the synthetic data two tests were performed. (1) Classify the URLs into malicious and benign to ensure the characteristics of the original data were preserved. (2) Use the Levenstein ratio to check the similarity between the real and synthetic URLs to ensure sufficient anonymization. The results from the classification test show that the synthetic data classifier only slightly underperformed the real data classifier; however, with having accuracy, precision, recall, sensitivity, and specificity above 99%, it can be concluded that the characteristics of the malicious and benign URLs were preserved. The Levenstein ratio tests showed a mean of 67% and 79% similarity for the benign and malicious URLs, respectively. In the end, the character-level LSTM model successfully generated an anonymized, synthetic dataset, that was characteristically similar to the original, which could pave the way for the publication of many more datasets in this way.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".