A View on Vulnerabilites: The Security Challenges of XAI (Academic Track)
Bibliographic record
Abstract
Modern deep learning methods have long been considered as black-boxes due to their opaque decision-making processes. Explainable Artificial Intelligence (XAI), however, has turned the tables: it provides insight into how these models work, promoting transparency that is crucial for accountability. Yet, recent developments in adversarial machine learning have highlighted vulnerabilities in XAI methods, raising concerns about security, reliability and trustworthiness, particularly in sensitive areas like healthcare and autonomous systems. Awareness of the potential risks associated with XAI is needed as its adoption increases, driven in part by the need to enhance compliance to regulations. This survey provides a holistic perspective on the security and safety landscape surrounding XAI, categorizing research on adversarial attacks against XAI and the misuse of explainability to enhance attacks on AI systems, such as evasion and privacy breaches. Our contribution includes identifying current insecurities in XAI and outlining future research directions in adversarial XAI. This work serves as an accessible foundation and outlook to recognize potential research gaps and define future directions. It identifies data modalities, such as time-series or graph data, and XAI methods that have not been extensively investigated for vulnerabilities in current research.
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 | 0.013 | 0.031 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.002 | 0.012 |
| Scholarly communication | 0.009 | 0.020 |
| Open science | 0.002 | 0.006 |
| Research integrity | 0.006 | 0.016 |
| Insufficient payload (model declined to judge) | 0.007 | 0.002 |
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".