XAI-Driven Security Systems: Enhancing Trust and Clarity in Automated Threat Detection and Response
Bibliographic record
Abstract
Cybersecurity threats are on an exponential rise thus necessitating the need to develop advanced level artificial intelligence (AI) systems to respond and detect threats automatically. But conventional artificial intelligence models in security operations centers (SOCs) have posed a major trust deficit between security analysts and significantly weakened the performance of the systems. This study proposes a full-scale XAI-based security model that incorporates the explainable artificial intelligence (XAI) methods with the sophisticated machine learning algorithms in providing superior visibility and user trust to the threat-detection autonomous systems. We combined SHAP and LIME explanation frameworks with ensemble models such as Random Forest, XGBoost, CNN and LSTM architectures to achieve an accuracy of 95% on the detection with an explanation fidelity score of more than 0.85. Based on such a thorough testing on benchmark datasets such as CIC-IDS2017 and UNSW-NB15 totaling to more than 5.3 million network flows, the system can deliver processing rates of maximum 10,000 events/sec per core with a maximum end-to-end latency of less than 100 ms. Researching 30 SOC analysts in a human-factors study, the scores on trust increased notably, with systems utilizing XAI reducing the amount of time needed to triage an alert by 34 percent and the analysts confidence of the system by 68 percent, according to a human-factors study utilizing XAIenhanced systems. The system satisfies serious requirements of compliance in future regulations such as the EU Artificial Intelligence Act 2024 and NIST AI Risk Management Framework, and gives auditable, explainable choices on security. The findings show that explainable AI integration into the processes of cybersecurity can improve false positive rates by as much as 22 percent without compromising on detection performance hence providing a new paradigm of an acceptable AI in security applications.
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 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 | 0.003 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, 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".