ML researcher spanning LLM-based social simulation, adversarial network security, and behavioral web mining.
7
papers here
7
reviewed by Evolve
4
first-author
1
corresponding
—
works · OpenAlex
3
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 data is available yet. This axis would measure publication volume and productivity — total papers, venue types, and annual output counts — which cannot be reliably scored from three 2026 arXiv preprints alone, so no estimate is made here.
Impact—
No impact data is available yet. This axis would measure citation counts, citations per paper, and uptake by other researchers; since the cataloged works are all recent 2026 preprints, no citation signal exists to score.
Focus0/100
Evolve assigns a focus score of 0, reflecting an almost perfectly even spread in the raw field distribution: each of three fields — LLM-based multi-agent simulation, adversarial network security, and web mining and user modeling — accounts for one third of cataloged output. With only three 2026 preprints, one per field, this zero indicates deliberate breadth rather than instability, and no dominant specialization has emerged yet.
Collaboration56/100
The collaboration score of 41 rests on a compact network: across the three cataloged 2026 papers, Berjawi has 7 distinct coauthors, with Rida Khatoun the only recurring name in the frequent-coauthor list, alongside Walid Fahs, Giuseppe Fenza, Makram Chehayeb, Amina Rizk, and Ralph Elsaghbini. Author positions rotate — first on the multi-agent simulation paper, middle on the web-browsing study, last on the DDoS work — suggesting flexible roles within small teams.
Activity81/100
No activity data is available yet. This axis would measure publishing cadence and recent output over time, which cannot be inferred from a single publication year (2026) in the catalog; no trend estimate is made here.
Reproducibility50/100
Evolve's reproducibility mean is 43 across 3 scored papers. The bias-amplification and behavioral-fingerprinting studies are rated medium readiness — the latter explicitly praised for transparent reporting — while the GAN-based DDoS framework is rated low, with ambiguous evaluation details and a simplified threat model cited. Releasing code, completing evaluation protocols, and statistically testing reported gains would most raise this score.
Research summary
Omran Berjawi (ORCID 0009-0008-6763-9630) is an early-career machine learning researcher whose public bibliographic footprint — resolved here from arXiv records in the Evolve catalog — currently comprises three 2026 preprints spread evenly across three areas: LLM-based multi-agent social simulation, adversarial machine learning for network security, and machine learning for web mining and user modeling. In the computational social science work, Berjawi is first author, showing that a biased LLM minority of at most 20% of agents can shift neutral agents beyond a Friedkin–Johnsen baseline while agent rhetoric decouples from numerical positions; Evolve judges the mechanism plausible but flags single-model and one-sided settings. On a GAN-based DDoS detection framework, where Berjawi is last author, WGAN-GP augmentation improves detector robustness under synthetic evasion tests, though the threat modeling and evaluation reporting draw critique. A third, middle-author paper reports roughly 90% user identification from short browsing sessions and about 80% next-domain prediction accuracy. Evolve's reproducibility assessment averages 43 across the three papers (medium–low readiness), and the collaboration score of 41 reflects a compact network of 7 distinct coauthors, including repeat collaborator Rida Khatoun. Output, impact, and activity metrics are not yet available; this profile is name-resolved from public bibliographic data and should be read as an early-stage profile.
Collaboration pattern
Across three 2026 arXiv preprints, Berjawi works in small teams totaling 7 distinct coauthors. Rida Khatoun is the only recurring name in the frequent-coauthor list, appearing twice, alongside Walid Fahs, Giuseppe Fenza, Makram Chehayeb, Amina Rizk, and Ralph Elsaghbini. Author positions rotate across first, middle, and last, indicating flexible roles within compact teams; the collaboration dimension scores 41 on this basis. No corresponding-author role appears in the catalog.
Research keywords
Highlights
First-author study: small biased LLM minorities bend multi-agent opinion dynamics
GAN-based DDoS detection hardened with WGAN-GP augmentation
~90% user identification from short web browsing sessions
Timeline
- 2026First-author arXiv preprint showing that biased LLM minorities amplify opinion shifts and decouple rhetoric in multi-agent networks.
- 2026Last-author arXiv preprint proposing a GAN-based framework using WGAN-GP augmentation for robust DDoS attack detection.
- 2026Middle-author arXiv preprint reporting ~90% user identification and ~80% next-domain prediction from web browsing behavior.
Papers in the Evolve catalog
- first author
Digital Twins for Opinion Dynamics: A Generative LLM Framework for Social Networks
- first author
AURA: Adaptive Uncertainty-Routed Analysis for Email Threat Detection
Delphi Scanner: efficient and interpretable static malware detection via API sequence modeling
Behavioral Fingerprinting and Navigation Prediction in Web Browsing
A GAN-Based Framework for Robust DDoS Attack Detection
- first author
Bias Amplification in Multi-Agent Network: How Biased Agents Shape Opinions and Rhetoric
- first authorcorresponding
Digital Persuasion: Understanding the Impact of Online Influencers on Public Opinion
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 rest on only three 2026 arXiv preprints; output, impact, and activity metrics are missing from the underlying data.
- No institution or country metadata is available, so no affiliation-based claims are made in this profile.
- Reproducibility results (mean 43 over 3 papers) and review verdicts reflect Evolve's editorial assessment, not external replication.
- Paper metadata (titles, years, sources, authorship) comes from the Evolve catalog snapshot of arXiv records; no peer-reviewed venues are recorded.