Employing cluster-based class decomposition approach to detect phishing websites using machine learning classifiers
Bibliographic record
Abstract
Phishing is an attack by cybercriminals to obtain sensitive information such as account IDs, usernames, and passwords through the use of the anonymous structure of the Internet. Although software companies are launching new anti-phishing tools that use blacklists, heuristics, visual methods, and machine learning-based methods, these products cannot prevent all phishing attacks. This research offers an opportunity to increase accuracy in the detection of phishing sites. This study develops a model using machine learning algorithms, specifically the decision tree and the random forest, due to their outperforming the rest of the classifiers and being accredited by researchers in this field to achieve the highest accuracy. The study is based on two phases: the first phase is to measure the accuracy of classifiers on the dataset in the usual way before and after feature selection. The second phase uses the class decomposition approach and measures the accuracy of classifiers in the dataset before feature selection and after feature selection to detect phishing sites. The class decomposition approach is a technique to improve the performance of classifiers by distributing each class into clusters and renaming the examples of each cluster with a new class. This provides a specific metric that more accurately predicts the level of phishing. Testing on a dataset containing 11,055 instances, 4,898 phishing, and 6,157 legitimate, each instance has 30 features. It achieved the highest accuracy in the first phase through the random forest algorithm by 96.9% before feature selection, and after feature selection, it was by 97.1%. In the second phase, the highest accuracy of both the decision tree and random forest classifiers was achieved by 100% with the two and four classes after feature selection. While before feature selection, the random forest algorithm achieved 100% with only the two classes.
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 | 0.003 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.003 | 0.002 |
| 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".