Adversarial ML and privacy-preserving distributed learning researcher at Southern Illinois University Carbondale.
4
papers here
4
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
10
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
No output-dimension score is available yet. This axis would measure publication volume and productivity over time; the record currently lacks the complete, disambiguated publication set needed to compute it, so no score or estimate is offered here.
Impact—
No impact score is available yet. This axis would measure citation-based influence such as citation volume and field-normalized impact; only basic indicators (h-index 10, i10-index 11) are on record, which is insufficient to score the dimension.
Focus0/100
The focus score of 0 reflects a broad rather than concentrated agenda: the field distribution splits evenly (0.5/0.5) between AI security/adversarial machine learning and privacy-preserving ML for mental health, and the topic profile spans five clusters — Blockchain Technology Applications and Security (9), Privacy-Preserving Technologies in Data (6), UAV Applications and Optimization (6), Adversarial Robustness in Machine Learning (6), and Robotic Path Planning Algorithms (4).
Collaboration28/100
Collaboration scores 28, consistent with a compact working circle: only 3 distinct co-authors appear in the record, with Md Khalid Syfullah and Abdullahil Kafi recurring as the main partners. On both Evolve-catalog preprints the author sits in the last position without corresponding-author status, suggesting a contributor role within a small, stable team rather than a broad multi-institution network.
Activity70/100
Activity scores 70, the strongest axis in the radar. OpenAlex records 10 works in the last three years, and the countsByYear series stays active throughout: 8 (2021), 3 (2022), 6 (2023), 2 (2024), 3 (2025), and 5 (2026), indicating sustained output with a dip in 2024 and renewed momentum in 2026.
Reproducibility38/100
Reproducibility scores 20, with a reproMean of 20 computed over 2 reviewed papers — both arXiv preprints rated low for reproducibility readiness. RAG-NAROK omits generator/judge models, detector calibration, and strong baselines, while the split-learning paper's privacy guarantee is limited to data locality and its metric (0.708 average ANLS) is described as lenient, so external replication currently takes extra effort.
Research summary
Alvi Ataur Khalil is a researcher at Southern Illinois University Carbondale (US) whose OpenAlex-resolved profile splits evenly — 0.5/0.5 in the field distribution — between AI security/adversarial machine learning and machine learning for mental health with privacy-preserving distributed learning. Bibliometric indicators show an h-index of 10 and an i10-index of 11, with 10 works in the last three years and yearly counts of 8 (2021), 3 (2022), 6 (2023), 2 (2024), 3 (2025), and 5 (2026). The topic profile spans five clusters, led by Blockchain Technology Applications and Security (9), with Privacy-Preserving Technologies in Data, UAV Applications and Optimization, and Adversarial Robustness in Machine Learning (6 each), and Robotic Path Planning Algorithms (4). Two 2026 arXiv preprints reviewed in the Evolve catalog — RAG-NAROK and a schema-aware split-learning framework for mental distress prediction — place the author in the last position on both. Both are rated low for reproducibility readiness, yielding a reproducibility mean of 20 across the 2 reviewed papers. Collaboration is narrow (3 distinct co-authors), with Md Khalid Syfullah and Abdullahil Kafi as recurring partners.
Collaboration pattern
Collaboration is concentrated: 3 distinct co-authors, with Md Khalid Syfullah and Abdullahil Kafi as recurring partners. Both Evolve-catalog preprints place the author last, without corresponding-author designation, consistent with a senior-contributor role in a small, stable team. The topic spread — blockchain security, privacy-preserving learning, UAV optimization, adversarial robustness, and robotic path planning — suggests the same compact group works across application domains rather than forming external ties. Note the source frequent-coauthor list includes the subject's own name.
Research keywords
Highlights
RAG-NAROK: context-adaptive knowledge corpus poisoning in retrieval-augmented generation
Schema-aware split learning for privacy-preserving mental distress prediction
Timeline
- 2021Most productive year in the OpenAlex countsByYear record, with 8 publications.
- 2023Output rebounds to 6 publications, the second-highest count in the record.
- 2025Three publications recorded, maintaining an active cadence going into 2026.
- 2026Two arXiv preprints enter the Evolve catalog — RAG-NAROK on RAG corpus poisoning and a schema-aware split-learning framework for mental distress prediction — in a year with 5 works recorded overall.
Papers in the Evolve catalog
RAG-NAROK: Retrieval-Aware Knowledge Corpus Poisoning in RAG with Source-specific Refutation
LLM-Based Schema-Aware Split Learning for Privacy-Preserving Mental Distress Prediction Across Heterogeneous Surveys
Don't Send What You Don't Need: Question-Guided Token Pruning as a Privacy Defense for Vision-Language Models
USPLIT-VQA: U-Shaped Split Learning for Visual Question Answering with Contribution-Aware Weighted Aggregation
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometrics are derived from OpenAlex, with author identity resolved via OpenAlex disambiguation; no other bibliographic sources were used.
- Output and impact dimension scores are not computed; citation-level data are insufficient, and no estimates are provided.
- In-depth Evolve review covers only 2 arXiv preprints (both 2026); the reproducibility mean of 20 is based on these 2 early-stage preprints only.
- The source frequent-coauthor list contains the subject's own name, so co-author statistics may include a self-occurrence artifact.