Improving malicious URL re-evaluation scheduling through an empirical study of malware download centers
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
The retrieval and analysis of malicious content is an essential task for security researchers. At the same time, the distributors of malicious files deploy countermeasures to evade the scrutiny of security researchers. This paper investigates two techniques used by malware download centers: frequently updating the malicious payload, and blacklisting (i.e., refusing HTTP requests from researchers based on their IP). To this end, we sent HTTP requests to malware download centers over a period of four months. The requests are distributed across two pools of IPs, one exhibiting high volume research behaviour and another exhibiting semi-random, low volume behaviour. We identify several distinct update patterns, including sites that do not update the binary at all, sites that update the binary for each new client but then repeatedly serve a specific binary to the same client, sites that periodically update the binary with periods ranging from one hour to 84 days, and server-side polymorphic sites, that deliver new binaries for each HTTP request. From this classification we identify several guidelines for crawlers that re-query malware download centers looking for binary updates. We propose a scheduling algorithm that incorporates these guidelines, and perform a limited evaluation of the algorithm using the data we collected. We analyze our data for evidence of blacklisting and find strong evidence that a small minority of URLs blacklisted our high volume IPs, but for the majority of malicious URLs studied, there was no observable blacklisting response, despite issuing over over 1.5 million requests to 5001 different malware download centers.
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 | 0.000 | 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 it