Edward Raff
—
Applied ML researcher across compression-based learning, cybersecurity ML, and deployed SaaS abuse detection
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-14) 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 publication-volume data is in scope: the catalog holds only the three Evolve-reviewed 2026 arXiv papers, so total output, venue history, and publication cadence cannot be measured yet.
Impact—
Citation metrics such as citation counts and h-index are absent from this snapshot, so research impact cannot be scored; the axis would measure citations and scholarly influence of the reviewed work.
Focus0/100
Evolve's focus score is 0, and the raw field distribution explains why: the three cataloged papers divide perfectly evenly at about 33% each across cybersecurity/applied ML (SaaS abuse), ML for cybersecurity/software engineering, and ML for information theory/data compression. This is breadth across problem domains rather than depth in one niche, with compression ideas also surfacing in the security work.
Collaboration50/100
The collaboration score is 50 with 11 distinct coauthors across the three cataloged papers. Recurring partners include Michael Brautbar, Charles K. Nicholas, Vitaly Zaytsev, and John Hurwitz, alongside a larger group (Shaefer Drew, Paul Knight, Lana Peric-McDermott, Simran Sarin, Nickolas Machado, Hanna Albright, Manu Nandan) suggesting team-based industry work; he sits mid-author on two papers and last on one, never corresponding.
Activity—
Recent-activity data such as per-year publication counts is unavailable beyond the three 2026 papers, so current research tempo and sustained output cannot be assessed.
Reproducibility43/100
Evolve's reproducibility mean is 43 over the three reviewed papers, with readiness rated medium for the compression-based learning survey and the file type identification paper, and low for the CrowdStrike abuse-detection paper, where proprietary deployment data limits replication. Review critiques of missing significance testing and missing ablations directly depress this score even where reported metrics are detailed.
Research summary
Edward Raff is an applied machine learning researcher whose recent work splits roughly evenly across three strands: machine learning for cybersecurity, anomaly detection for SaaS platforms, and the information-theoretic foundations of compression-based learning. All three papers in the Evolve catalog are 2026 arXiv preprints. The compression-based learning introduction formalizes the field as a four-choice design framework spanning lossless compression, Normalized Compression Distance, Minimum Description Length, and Kolmogorov complexity; Evolve judges its experiments competitive and notably strong on malware, though lacking significance testing. A second paper reports LABELER, a convolutional file type identifier that beats Magika across 54 text file types (Macro F1 0.98 vs 0.91, 3.7x faster), with Evolve cautioning that the gains look confounded and the evaluation lacks ablations. The third, a CrowdStrike collaboration, documents a deployed abuse-detection system pairing Isolation Forest with expert risk scores, achieving 35% coverage and 30-58% fewer alerts on small-sample evidence. Evolve scores reproducibility at 43 across three papers and collaboration at 50, with 11 distinct coauthors; he appears as a middle author on two papers and last author on one. The profile is name-resolved from public bibliographic data, with no institutional affiliation or ORCID attached, and output, impact, and activity metrics remain unavailable.
Collaboration pattern
Across the three cataloged papers, Edward Raff works with 11 distinct coauthors (collaboration score 50). The roster — Michael Brautbar, Charles K. Nicholas, Vitaly Zaytsev, John Hurwitz, Shaefer Drew, Paul Knight, Lana Peric-McDermott, Simran Sarin, Nickolas Machado, Hanna Albright, and Manu Nandan — points to team-authored, industry-adjacent projects rather than solo work. He appears as a middle author on two papers and last author on one, with no corresponding-author role recorded.
Research keywords
Highlights
Formalizing compression-based machine learning as a design framework
LABELER tops Magika on text file type identification
Deployed ML abuse detection in a security platform
Timeline
- 2026Details a deployed SaaS abuse-detection system with CrowdStrike, combining Isolation Forest and expert risk scores (35% coverage, 30-58% fewer alerts)
- 2026Reports LABELER, a fast CNN-based text file type identifier benchmarked against Magika on 54 text file types
- 2026Publishes an introduction to compression-based machine learning spanning NCD, MDL, and Kolmogorov complexity
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 small bibliometric snapshot (OpenAlex-derived); only the three Evolve-reviewed 2026 arXiv papers are in scope.
- No institutional affiliation, country, or ORCID is attached; the profile is name-resolved from public bibliographic data and other researchers named Edward Raff are not distinguished.
- Output, impact, and activity dimensions lack underlying data (no citation counts or per-year publication volumes), so they are left unscored.
- All cataloged papers are 2026 arXiv preprints; no peer-reviewed venue, grant, or award information is included.