Joakim Loxdal

KTH Royal Institute of Technology

Unclaimed · AI-structured profileORCID OpenAlex

KTH researcher building simulation and cyber-range testbeds for reinforcement-learning-driven autonomous cyber defense

papers here

reviewed by Evolve

first-author

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—

Not yet scored: publication-volume data is insufficient in the current snapshot. This axis would measure the count and cadence of indexed outputs; only the two 2026 arXiv preprints in the Evolve catalog can be counted today.

Impact—

Not yet scored: citation and influence metrics are unavailable in the underlying data. This axis would measure citations, usage, and uptake of the work; both cataloged papers are recent 2026 preprints, so no reliable impact signal exists yet.

Focus0/100

The focus axis scores 0 while the field distribution splits evenly — 0.5 for computer security (AI-driven cyber operations simulation) and 0.5 for cybersecurity (autonomous cyber defense). Read together, the cataloged work spans two tightly adjacent subfields rather than a single niche: both papers address simulating attacks and defenses, one as a graph-based simulator, the other as a physical cyber-range evaluation.

Collaboration39/100

The collaboration dimension scores 39, based on 6 distinct coauthors in the underlying data. Both cataloged 2026 papers list Joakim Loxdal in a middle author position without corresponding-author role, consistent with contribution inside an established KTH team. Recurring coauthors include Jakob Nyberg, Andrei Buhaiu, Pontus Johnson, Mathias Ekstedt, Sandor Berglund, and Teodor Sommestad, indicating a compact, stable group rather than a wide external network.

Activity—

Not yet scored: year-by-year publication cadence is insufficient in the current snapshot. This axis would measure publishing consistency and recent output rate; the reviewed record contains only 2026-dated preprints, so no multi-year activity trend can be computed.

Reproducibility55/100

Reproducibility averages 55 across the 2 papers scored. Evolve rates both as medium readiness: the MAL Simulator case-study reinforcement-learning results are credible but validated only in simulation, while the cyber-range study reports agents beating a heuristic across a 37-VM testbed with thin statistics and a NoOp baseline winning under the Guided attacker. Artifacts appear implementable yet would need stronger experimental controls.

Research summary

Joakim Loxdal is a researcher at KTH Royal Institute of Technology working at the intersection of cyber operations modeling and autonomous cyber defense. The reviewed record comprises two 2026 arXiv preprints, both with him in a middle author position. 'The MAL Simulator: Cyber Operations Simulation based on Attack & Defense Graphs' extends the Meta Attack Language ecosystem into a simulator with language-driven adaptability and dual attack/defense interfaces; Evolve's review finds its case-study reinforcement-learning results credible but validated only in simulation. 'A Cyber Range Evaluation of Autonomous Network Incident Response Agents' moves evaluation onto a 37-VM cyber range and reports reinforcement-learning agents beating a heuristic baseline, with thin statistics and a NoOp policy winning under the Guided attacker noted as caveats. The field distribution splits evenly — 0.5 computer security (AI-driven cyber operations simulation), 0.5 cybersecurity (autonomous cyber defense) — and the focus score is 0, reflecting coverage of two adjacent subfields rather than one niche. Collaboration scores 39 on 6 distinct coauthors, a compact KTH-centered group including Jakob Nyberg, Andrei Buhaiu, Pontus Johnson, Mathias Ekstedt, Sandor Berglund, and Teodor Sommestad. Reproducibility averages 55 across 2 scored papers, both medium readiness. Output, impact, and activity axes lack sufficient data and remain unscored, so this should be read as an early-stage, simulation-focused profile.

Collaboration pattern

Collaboration scores 39 from 6 distinct coauthors in the underlying data. On both cataloged 2026 papers Joakim Loxdal appears in a middle author position without corresponding-author role. The recurring coauthor set — Jakob Nyberg, Andrei Buhaiu, Pontus Johnson, Mathias Ekstedt, Sandor Berglund, and Teodor Sommestad — points to a compact, stable KTH-based team rather than a broad external network; no cross-institution pattern is visible in the reviewed record.

Research keywords

autonomous cyber defensecyber operations simulationattack graphsMeta Attack Language (MAL)reinforcement learningcyber rangesim-to-real transfernetwork incident responseattack simulationautomated penetration testing

Highlights

MAL Simulator brings language-driven adaptability to attack/defense graph simulation
Reviewed on arXiv (2026), this work extends the Meta Attack Language ecosystem into a cyber operations simulator with dual attack and defense interfaces. Evolve's verdict: the case-study reinforcement-learning results are credible, but validated only in simulation; reproducibility readiness is rated medium.
37-VM cyber-range test shows RL agents beating a heuristic — with caveats
In this 2026 arXiv study, autonomous network incident response agents were evaluated on a 37-VM cyber range, forming a credible sim-to-real transfer study. Evolve notes the reinforcement-learning agents beat a heuristic baseline, but statistics are thin and a NoOp policy wins under the Guided attacker; reproducibility readiness is medium.

Timeline

2026Co-authored 'The MAL Simulator: Cyber Operations Simulation based on Attack & Defense Graphs' (arXiv preprint; middle author position).
2026Co-authored 'A Cyber Range Evaluation of Autonomous Network Incident Response Agents' (arXiv preprint; middle author position).
2026Both cataloged outputs fall within 2026, which stands as the earliest — and only — year in this profile's reviewed record.

Papers in the Evolve catalog

Frequent co-authors

Jakob Nyberg×2 Andrei Buhaiu×2 Pontus Johnson×2 Mathias Ekstedt×2 Sandor Berglund×1 Teodor Sommestad×1

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 dimensions are computed from OpenAlex-derived bibliometric data only; output, impact, and activity axes lack sufficient data and are unscored.
The reviewed record covers only two arXiv preprints dated 2026; no peer-reviewed venue or citation data is included.
Country is not present in the source data; affiliation is listed as KTH Royal Institute of Technology and identity is name-resolved via ORCID 0009-0001-3488-2246.
The reproducibility mean (55) rests on only 2 scored papers, both rated medium readiness in Evolve review, so the estimate is narrow.