Bibliographic record
Abstract
Everyday, security experts face a growing number of security events that affecting people well-being, their information systems and sometimes the critical infrastructure. The sooner they can detect and understand these threats, the more they can mitigate and forensically investigate them. Therefore, they need to have a situation awareness of the existing security events and their possible effects. However, given the large number of events, it can be difficult for security analysts and researchers to handle this flow of information in an adequate manner and answer the following questions in near-real time: what are the current security events? How long do they last? In this paper, we will try to answer these issues by leveraging social networks that contain a massive amount of valuable information on many topics. However, because of the very high volume, extracting meaningful information can be challenging. For this reason, we propose SONAR: an automatic, self-learned framework that can detect, geolocate and categorize cyber security events in near-real time over the Twitter stream. SONAR is based on a taxonomy of cyber security events and a set of seed keywords describing type of events that we want to follow in order to start detecting events. Using these seed keywords, it automatically discovers new relevant keywords such as malware names to enhance the range of detection while staying in the same domain. Using a custom taxonomy describing all type of cyber threats, we demonstrate the capabilities of SONAR on a dataset of approximately 47.8 million tweets related to cyber security in the last 9 months. SONAR could efficiently and effectively detect, categorize and monitor cyber security related events before getting on the security news, and it could automatically discover new security terminologies with their event. Additionally, SONAR is highly scalable and customizable by design; therefore we could adapt SONAR framework for virtually any type of events that experts are interested in.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.126 | 0.167 |
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 source (direct Gemma or distilled Codex), 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".