MiniWarner: An Novel and Automatic Malicious Phishing Mini-apps Detection Approach
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
WeChat mini-apps are “sub-applications” built within the WeChat platform. Unlike full-function native applications, they are streamlined, “light” versions of the apps, and enable users to open and use them inside WeChat without downloading and installation. Since being introduced by WeChat in 2017, 4.3 million WeChat mini-programs have been developed, and they attract around 410 million daily active users Up to 2021. However, motivated by financial gains, many malicious mini-app developers use some intended description and icon to mislead users to click and open their mini-apps. These mini-apps are full of annoying advertisements and collect users’ privacy information stealthily, which can expose users to privacy risks and financial losses. Although security personnel of WeChat has enforced various countermeasures to prevent malicious phishing mini-apps sneaking into WeChat, rampant malicious leading mini-apps still have been observed recently. In this paper, we present MiniWarner, a novel approach that leverages Natural Language Processing and a number of reverse engineering techniques to detect whether a mini-app is malicious and phishing when users open it. MiniWarner will only ask users whether to continue to open the malicious phishing mini-app, thus it can protect users against the intended misleading by attackers, and still preserve the original user experience. Besides, this approach is implemented as an Xposed module, making it practical to be quickly deployed on a large number of user devices. Our paper will introduce how we developed MiniWarner and the measurement results of MiniWarner in detail.
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.
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 | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.015 |
| Open science | 0.001 | 0.001 |
| 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 it