AI security governance and federated learning researcher at SBA Research; first author on agent incident reporting.
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
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—
Output cannot be scored yet: the current snapshot provides no publication-count or venue data for this author. This axis would measure research productivity, such as papers per year, venue types, and first/corresponding-author share across the record.
Impact—
Impact cannot be scored yet: no citation counts, h-index, or downstream uptake are available in the supplied data. This axis would measure scholarly influence, including citations to the two catalogued papers and adoption of the proposed reporting frameworks.
Focus0/100
Focus scores 0 because the record is split exactly evenly across two fields: AI security and governance (0.5) and medical informatics built on federated learning infrastructure (0.5). With only two catalogued papers, one per field, this reads as a deliberate dual track — agent-facing security governance on one side, clinical data infrastructure on the other — rather than unfocused work, but the sample is too small to say firmly.
Collaboration84/100
Collaboration scores 84, backed by 67 distinct coauthors in the record — a wide network spanning both of her fields. The frequent-coauthor list includes Eugene Bagdasarian, Luca Beurer-Kellner, Battista Biggio, Heather Frase, Daphne Ippolito, and Somesh Jha, indicating active ties across AI security and federated clinical data communities. Her two catalogued papers also show role flexibility: first author on one, middle author on the other.
Activity—
Activity cannot be scored yet: the snapshot contains no per-year publication counts or cadence measures. This axis would measure research tempo, such as outputs per year and recency of contributions; only the two 2026 arXiv papers are available as evidence.
Reproducibility38/100
Reproducibility scores 38 (reproMean) across the 2 papers Evolve has assessed. The first-authored agent incident-reporting position paper is rated low readiness, lacking a solicitation method or validation; FL-Net is rated medium, with strong systems integration but simulated, self-scored evaluation. Both papers fall short of the conditions needed for independent replication, so artifact release and external validation would lift this axis.
Research summary
Anastasia Pustozerova is a researcher at SBA Research whose catalogued record splits exactly evenly between two fields: AI security and governance (0.5) and medical informatics built on federated learning infrastructure (0.5). Evolve's catalog includes two of her 2026 arXiv papers. The first, with her as first author, is 'Beyond Predictable Paths: Redefining AI Security Incident Reporting for Agents,' an expert position paper arguing that incident reporting must extend beyond predictable software paths to capture agent trajectories, autonomy, delegation and identity, and multi-agent data; Evolve finds the framework clear but notes it lacks a solicitation method or validation, with low reproducibility readiness. The second, a middle-author contribution, presents FL-Net, a one-stop federated learning platform for multi-center medical data mining that claims to meet all five clinical-FL requirements; Evolve credits strong systems integration while flagging simulated, self-scored evaluation and medium reproducibility readiness. Her collaboration footprint is broad — 67 distinct coauthors, a collaboration score of 84, and frequent coauthors including Eugene Bagdasarian, Luca Beurer-Kellner, Battista Biggio, Daphne Ippolito, and Somesh Jha. Reproducibility averages 38 across the two reviewed papers. Output, impact, and activity cannot be scored because the snapshot lacks citation and publication-count data; the profile is name-resolved from public bibliographic data.
Collaboration pattern
Her collaboration footprint is unusually broad for a two-paper catalogued record: 67 distinct coauthors and a collaboration score of 84. The frequent-coauthor list spans both of her fields, including Eugene Bagdasarian, Luca Beurer-Kellner, Battista Biggio, Somesh Jha, Daphne Ippolito, Heather Frase, Marc Fischer, Sean McGregor, Nico Ebert, David Filip, David Hofer, and Juliane Hoffmann. She contributes as first author on the agent incident-reporting paper and as a middle author on FL-Net.
Research keywords
Highlights
First-author position paper redefines incident reporting for AI agents
FL-Net: a one-stop federated learning platform for multi-center medical data
Dual-track record: AI agent security governance alongside clinical FL infrastructure
Her catalogued record splits evenly between AI security and governance (0.5) and medical informatics on federated learning infrastructure (0.5), with 67 distinct coauthors connecting the two tracks. The pattern pairs governance-oriented position work with deployable clinical systems rather than a single narrow specialty.
Timeline
- 2026First author of 'Beyond Predictable Paths: Redefining AI Security Incident Reporting for Agents' (arXiv), an expert position paper on incident reporting for AI agents.
- 2026Middle author on 'Multi-center Medical Data Mining with FL-Net' (arXiv), a one-stop federated learning platform for multi-center medical data mining.
- 2026Catalogued 2026 record spans two fields evenly — AI security and governance and medical informatics FL infrastructure — with 67 distinct coauthors in total.
Papers in the Evolve catalog
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.
- Dimension scores are computed from public bibliometric data (ORCID 0000-0002-8197-1145) and are not independently verified; the profile is name-resolved from this record.
- Only the two 2026 arXiv papers above carry Evolve reviews; any other outputs by this author are not assessed in this profile.
- Output, impact, and activity axes lack underlying citation and publication-count data and are deliberately left unscored.
- The focus score of 0 reflects a 50/50 split across two fields within a two-paper sample and may overstate fragmentation.