Security researcher at Minjiang University measuring and mediating privacy leakage in LLM-agent systems
3
papers here
3
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
0
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—
Insufficient data: the OpenAlex record shows h-index 0 and i10-index 0 with only recent-year counts, so output cannot be meaningfully scored. This axis would measure cumulative publication volume, venue quality, and productivity trends over time.
Impact—
Insufficient data: citation signals (h-index 0, i10-index 0) are too thin to score impact. This axis would measure citation uptake, influence on follow-up work, and recognition across the LLM-agent privacy community.
Focus0/100
The focus score is 0, derived from a field distribution that splits evenly at 0.33 each across AI security and privacy evaluation, AI security and privacy, and computer security for LLM-agent systems privacy. In practice the three labels describe one theme, and OpenAlex topics cluster on privacy-preserving technologies (4), privacy, security, and data protection (4), access control and trust (2), and adversarial robustness (2) — so the low score reflects label fragmentation more than scattered research.
Collaboration56/100
The OpenAlex record lists 16 distinct coauthors in the recent window, and the collaboration score is 56. Recurring names include Tao Huang and Chen Hou, with Chen Hou appearing three times in the frequent-coauthor list, alongside Guolong Zheng, Huizhen Huang, Guoxiong Long, Jiayang Meng, Xu Yang, and Xuechao Yang. All three reviewed papers are multi-authored: Wu is first author on ASLEval and a middle author on CIPL and PAPC, a team-based pattern rather than solo work.
Activity47/100
Activity scores 47, based on a thin recent record: 4 works in the last three years, all counted in 2026. All three Evolve-reviewed papers are 2026 arXiv preprints, suggesting a burst of recent output in LLM-agent privacy with no longer publication history captured in this record — early-career or newly name-resolved activity rather than a sustained multi-year cadence.
Reproducibility55/100
Reproducibility scores 55, the mean readiness across the 3 Evolve-reviewed papers, each rated medium. Evolve's verdicts note coherent results across 5 LLM providers in CIPL but no statistical testing, a rigorous protocol in ASLEval limited by synthetic environments and small samples, and partly-by-construction zero-leak results in PAPC — methodological care with acknowledged evaluation gaps.
Research summary
Guosen Wu is a researcher at Minjiang University, China, whose work as captured in this OpenAlex-resolved profile centers on privacy in LLM-agent systems. The Evolve catalog contains three of Wu's 2026 arXiv papers, all addressing how large language model agents leak, propagate, and can be prevented from exposing private information. ASLEval, with Wu as first author, measures privacy-exposure displacement in LLM-agent sessions and finds local privacy proxies miss roughly 47% of session-level exposure; CIPL proposes a channel-aware framework for recoverable privacy leakage, with results coherent across five LLM providers; and PAPC introduces event-level platform mediation for privacy-propagation externalities in multi-agent workflows. OpenAlex topics concentrate on privacy-preserving technologies and privacy, security, and data protection (4 works each), with secondary coverage of access control and trust and adversarial robustness in machine learning (2 each). The record shows 4 works in the last three years, all in 2026, 16 distinct coauthors, a collaboration score of 56, activity of 47, and reproducibility of 55, with medium readiness across all three reviewed papers. Evolve's verdicts consistently credit rigorous framing while flagging evaluation gaps: missing statistical testing, synthetic environments, small samples, and partly-by-construction results. Output and impact cannot be scored from the current record, making this an early-stage profile for an evidently active but not yet citation-visible researcher.
Collaboration pattern
The record lists 16 distinct coauthors in the recent window (collaboration score 56). Tao Huang and Chen Hou recur most often, with Chen Hou appearing three times in the frequent-coauthor list, followed by Guolong Zheng, Huizhen Huang, Guoxiong Long, Jiayang Meng, Xu Yang, and Xuechao Yang. All three reviewed papers are team-authored: Wu is first author on ASLEval and a middle author on CIPL and PAPC, indicating a group-based working pattern rather than solo scholarship.
Research keywords
Highlights
ASLEval: first-author benchmark finds local privacy proxies miss ~47% of LLM-agent session exposure
CIPL: channel-aware framing of recoverable privacy leakage in LLM agents
PAPC: event-level platform mediation for privacy propagation in multi-agent workflows
Timeline
- 2026Publishes ASLEval on arXiv as first author, measuring privacy-exposure displacement in LLM-agent sessions and reporting that local privacy proxies miss ~47% of session-level exposure.
- 2026Co-authors PAPC on arXiv, proposing event-level platform mediation for privacy-propagation externalities in multi-agent LLM workflows.
- 2026Co-authors CIPL on arXiv, a channel-aware framework formalizing recoverable privacy leakage in LLM agents across 5 providers.
- 2026OpenAlex records all 4 works from the recent three-year window in 2026, clustered in privacy and security topic areas.
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.
- Bibliometric scores come from a single OpenAlex record; output and impact lack sufficient underlying data and are not scored.
- Identity is name-resolved from public bibliographic data; no ORCID was available, and 'Guosen Wu' is a common Chinese name.
- All three reviewed papers are 2026 arXiv preprints; no peer-reviewed venue, citation, or career-history information is in scope.
- Reproducibility (mean 55) reflects Evolve review of only three papers; activity and collaboration counts cover a limited recent window.