Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
A beginner machine-learning experiment reported 96.70% test accuracy for a Random Forest trained to distinguish phishing websites from legitimate ones. That result belongs to one model, one held-out split, and a dataset donated in 2015—it is not evidence that the detector can identify today’s phishing sites reliably in the wild.
What the 96.70% result means
In a 2026 DEV Community post, ELNAZEER DAWOD describes training three classifiers on the UCI Phishing Websites dataset. The author reports Random Forest as the strongest of the three on the experiment’s test set, with SVM and Logistic Regression behind it.
| Classifier | Reported test accuracy |
|---|---|
| Random Forest | 96.70% (ELNAZEER DAWOD, 2026) |
| Support Vector Machine (SVM) | 94.71% (ELNAZEER DAWOD, 2026) |
| Logistic Regression | 92.45% (ELNAZEER DAWOD, 2026) |
The figures are the author’s reported test accuracies, not independently verified measurements. Accuracy is the share of test examples classified correctly. It does not reveal how many phishing pages were missed or how many legitimate pages were incorrectly flagged, so the highest accuracy alone does not establish which model would be safest or most useful in practice.
What data the experiment used
The UCI Machine Learning Repository record describes 11,055 website instances with 30 integer features. It credits Rami Mohammad and Lee McCluskey and records the dataset’s donation date as March 25, 2015. UCI Machine Learning Repository: Phishing Websites
The author says the experiment used an 80/20 train/test split and five-fold cross-validation. A held-out test split gives a model examples not used for fitting, while cross-validation evaluates performance across multiple partitions of the training data. These procedures provide useful checks on a dataset, but they do not show whether a model trained on older examples will generalize to newer phishing campaigns.
That distinction matters because the dataset is from 2015. UCI itself notes that reliable training data is a challenge and that the literature does not agree on a definitive set of features that characterizes phishing webpages. The reported score should therefore be read as a result on this particular labeled collection, rather than a current estimate of phishing-detection performance.
Which website features stood out
The author reports that SSL certificate state and anchor-link behavior ranked as the most important features in the model. The proposed explanation is that a fake domain may lack a valid certificate, while a copied page may retain links pointing to the legitimate site. The post explicitly frames this as interpretation rather than a causal result: “This is my interpretation of the result, not something the experiment proved.”
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Feature importance indicates which inputs the model relied on within this experiment; it does not prove that either feature causes a page to be phishing. Nor does the result establish that either signal will remain reliable across sites, time periods, or different data.
Rank #3
- 【Tired of constantly searching for or resetting your passwords?】 MOSA BEAR password keeper book is the perfect solution for you! This password book provides a dedicated place to securely store all your important website addresses, emails, usernames and passwords, ensuring your information is protected and easy to find. The well-designed log pages help you manage multiple accounts in a systematic way, saying goodbye to password confusion.
- 【Premium Design & Password Security】 The password book with alphabetical tabs features an anonymous cover design with no title on the cover, effectively avoiding information exposure. The password keeper design is specifically designed with password security in mind, providing space to record password hints instead of writing directly on the password itself, further protecting your important information.
- 【Simple Layout and Plenty of Space】The 160-page password logbook is designed to provide ample space to record passwords and other important information. It can store up to 414 passwords. In addition, it provides extra pages to record other information, such as email setup, card information, computer operating system information, software licenses, and more. The journal also includes 3 blank pages at the end for you to add additional notes.
- 【Palm-sized Size & Premium Quality】 This password notebook has an ideal size, 4.3" x 5.7", for carrying around, whether in a purse or pocket. Its sturdy glue binding allows the notebook to unfold smoothly and is more comfortable to use. The inner pages are made of high-quality 100GSM thick paper, which can effectively reduce ink penetration and ensure a cleaner and neater writing effect. The overall design takes into account both portability and durability, making it an ideal choice for recording important passwords.
- 【A-Z Tabs for Quick Search 】Our password book comes with alphabetical tabs to help you find the password you need quickly and easily. Alphabetically organized tabs ensure that you can quickly flip to the right section, saving you the time and hassle of searching for your password.
What would be needed to judge a real detector
For a practical security tool, overall accuracy is only one part of the evaluation. A detector’s errors can have different costs: missing a phishing page may expose a user to fraud, while flagging a legitimate site can block normal access. A fuller evaluation would report:
- Precision: among sites flagged as phishing, how many actually are phishing.
- Recall: among phishing sites in the test data, how many the model catches.
- False-positive and false-negative counts: the number of legitimate sites incorrectly flagged and phishing sites missed.
- Class balance: how many examples in each category are present, which affects how accuracy should be interpreted.
- External or time-based validation: performance on a separate, newer collection of sites, rather than only on a split of the same dataset.
The DEV post’s reviewed body does not provide precision, recall, a confusion matrix, or external and temporal validation results. Without them, readers cannot determine the experiment’s error profile or infer that the model is ready to protect users.
Rank #4
- Bookbound planner helps you keep track of passwords and favorite websites
- Room for over 200 entries; 3.5 x 6 inch page sizes
- User name and security questions field
- Tips for what makes a strong password; web resources; notes pages
- Printed on quality paper containing 30% post-consumer waste; black simulated leather cover; 3.63 x 6.13 x .21 inches
What the project does—and does not—show
The post presents a beginner’s machine-learning project using Python, scikit-learn, and a public dataset. Its value is as an illustration of training and comparing classifiers on labeled website features. The reported results do not establish a deployed security product, a guarantee against phishing, or a need for paid software or specialized hardware.
For the experiment itself, Random Forest had the top reported test accuracy among the three models. For real-world protection, the available figures are not enough to choose or validate a detector: current, independent evaluation and error metrics would be needed.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
