Enhanced Malware Prediction and Containment Using Bayesian Neural Networks
Bibliographic record
Abstract
In this paper, we present an integrated framework leveraging natural language processing (NLP) techniques and machine learning (ML) algorithms to detect malware at its early stage and predict its upcoming actions. We analyze application programming interface (API) call sequences in the same way as natural language inputs. Specifically, the proposed model employs Bi-LSTM neural networks and Bayesian neural networks (BNN) for this analysis. In the first part, a Bagging-XGBoost algorithm interprets consecutive API calls as 2-gram and 3-gram strings for early-stage malware detection and feature importance analysis. Additionally, a Bi-LSTM predicts the upcoming actions of an active malware by estimating the next API call in a sequence. Two separate Bayesian Bi-LSTMs are then developed in the second part to complement the above analysis. The first architecture is for early-stage malware detection, and the other is to predict the following action of active malware. The BNN not only predicts future malware actions but also assesses the uncertainty of each prediction. It enhances the process by providing the second and third most probable predictions, increasing system reliability and effectiveness. Our unified framework demonstrates efficiency in malware detection and action prediction, marking a significant advancement in countering malware threats. The Bayesian Bi-LSTM developed for predicting the next API call has an average accuracy of 89.53%. Additionally, the accuracy of the framework for malware detection at the early stage is 96.44%, demonstrating the superior performance of the proposed framework.
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 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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".