EXPLOITING AND SECURING MACHINE LEARNING: A CYBERSECURITY PERSPECTIVE ON ADVERSARIAL VULNERABILITIES AND COUNTERMEASURES
Bibliographic record
Abstract
Machine learning (ML) is now the new focus of cybersecurity, as it can be used to perform automated intrusion detection, malware classification, and anomaly recognition. ML models are extremely susceptible to adversarial perturbations, or small, intentional manipulation of data that can severely misclassify the input data but leave no discernible effects on the input data. The research discusses the vulnerabilities of the ML models to adversarial attacks and examines effective and reproducible countermeasures that enhance capacity against cybersecurity-related apps. The performance drops because of adversarial attacks, and determines the performance of adversarial retraining in causing model resilience against adversarial attacks. An open-source adversarial dataset was used in a semi-empirical experiment utilizing Fast Gradient Sign Method (FGSM). To evaluate the performance of a feed-forward neural network under clean, adversarial and post-defence settings, Python-based frameworks were trained, attacked and re-trained. The FGSM attack decreased model accuracy by about 19% which validates that adversarial noise is highly vulnerable. The model regained about 14% of the lost performance, and this enhanced the classification stability and detection accuracy after retraining on combined clean and perturbed data. The results indicate that reproducible experiments of lightweight can be used effectively to test adversarial threats. Although the retraining approaches are not complex, even basic methods lead to a substantial enhancement of ML resilience and facilitate reliable and trusted AI-determined cybersecurity infrastructures.
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 | 0.003 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.002 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 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 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".