Mohamed Chahine Ghanem

—

Unclaimed · AI-structured profileORCID OpenAlex

AI assurance and offensive-security researcher: independence-graded agentic-AI audits and LLM-driven pentest frameworks

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-09-27) 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 scorable: no publication counts or venue-volume data are available in the current snapshot. This axis would measure the volume and cadence of the researcher's publications, including preprints, over time.

Impact—

Not yet scorable: no citation counts, h-index, or venue-level influence data exist in this snapshot. This axis would measure citations, downstream uptake, and the researcher's influence on subsequent work by others.

Focus0/100

The radar focus score is 0, reflecting no single dominant field: the field distribution splits exactly 50/50 between AI governance and assurance (trustworthy AI/security) and cybersecurity — LLM-driven offensive security and agent systems. The two catalogued papers sit one per field — an agentic-AI audit protocol and an automated penetration-testing framework — so this reads as a deliberate dual profile around security-relevant, trustworthy AI rather than drift.

Collaboration36/100

The collaboration score is 32, computed from 4 distinct coauthors in the current snapshot — a compact pool pointing to small, repeat teams rather than a broad network. Recurring collaborators include Rushabh Vipulkumar Patel, Dipo Dunsin, and Mohammed Almaiah. The researcher also switches roles across projects: first and corresponding author on the audit-protocol paper, last author on PentestChain, showing both leadership and supporting participation.

Activity—

Not yet scorable: the only dated records are two 2026 papers, too little to establish a publishing rhythm. This axis would measure year-over-year publication cadence and sustained research engagement.

Reproducibility78/100

Reproducibility is the strongest measured axis, scoring 78 — the mean across 2 assessed papers. The independence-graded audit protocol rates high reproducibility readiness, though its validation is simulation-only with declared parameters; PentestChain rates medium, with a deterministic backbone but projected head-to-head and ablation evidence. Both Evolve verdicts explicitly declare evidence limits, a pattern suggesting the author documents scope honestly even where empirical validation is incomplete.

Research summary

Mohamed Chahine Ghanem (ORCID 0000-0002-7067-7848) works at the intersection of AI governance and assurance and LLM-driven offensive security. The current snapshot holds two 2026 arXiv papers, split evenly between his two fields (50/50 field distribution). As first and corresponding author he proposed an independence-graded audit protocol for agentic AI, which Evolve reviewers found formally grounded and honestly caveated while noting validation rests on simulation with declared parameters. As last author he contributed to PentestChain, a cost-aware, MCP-orchestrated framework for automated penetration testing with free-tier LLMs; its measured $0 cost applies to legacy targets only, with head-to-head and ablation evidence still projected. Reproducibility is his strongest measured dimension (mean 78 across 2 papers; readiness high and medium), consistent with a declared-parameter, deterministic-backbone working style. Collaboration is compact (score 32; 4 distinct coauthors), with recurring partners Rushabh Vipulkumar Patel, Dipo Dunsin, and Mohammed Almaiah, and visible role flexibility between leading and supporting positions. Publication volume, citations, and multi-year activity data are not yet available, so output, impact, and activity remain unscored; institution and country are likewise unrecorded.

Collaboration pattern

With only 4 distinct coauthors recorded, Mohamed Chahine Ghanem's collaboration footprint is compact and repeat-oriented. Rushabh Vipulkumar Patel, Dipo Dunsin, and Mohammed Almaiah recur as partners. Roles flex across projects: first and corresponding author leading the agentic-AI audit protocol, last author on the PentestChain framework — a pattern consistent with small teams whose members alternate between leading and supporting positions rather than forming large consortia.

Research keywords

agentic AIAI auditingauditor independencecommon-cause failurebeta-factor modelalgorithmic monocultureautomated penetration testinglarge language modelsModel Context Protocol (MCP)cost-aware AI cascadeLLM agentsdeterministic exploit map

Highlights

Independence-graded audit protocol for agentic AI
First and corresponding author on "Who Audits Whom, on What Substrate, with What Evidence?", which grades AI audits across three independence axes to guard against common-cause failure and algorithmic monoculture, drawing on a beta-factor model. Evolve reviewers judge it formally grounded and honestly caveated, though validated only by simulation with declared parameters (reproducibility readiness: high).
PentestChain: free-tier LLMs for automated pentesting
Last author on a cost-aware, MCP-orchestrated framework that chains free-tier LLM agents over a deterministic exploit map for automated penetration testing. Evolve notes the measured $0 cost covers legacy targets only, with head-to-head and ablation evidence still projected (reproducibility readiness: medium).
A declared-parameter, reproducibility-first working style
Across both papers, Evolve's reproducibility assessments average 78 (readiness high and medium), and both verdicts explicitly flag evidence limits — simulation-only validation and projected comparisons — indicating a habit of declaring scope and parameters rather than overstating results.

Timeline

2026Published on arXiv as first and corresponding author an independence-graded audit protocol for agentic AI, addressing auditor independence, common-cause failure, and algorithmic monoculture.
2026Co-authored PentestChain on arXiv as last author: a cost-aware, MCP-orchestrated framework for automated penetration testing with free-tier LLMs.
2026Both papers entered the Evolve catalog with reproducibility readiness ratings of high and medium, averaging 78 across the two assessments.

Papers in the Evolve catalog

2026-09-24
ASIRF: An Agentic Framework for Context-Dependent Sensitive Information Redaction
2026-09-17first authorcorresponding
Who Audits Whom, on What Substrate, with What Evidence? An Independence-Graded Audit Protocol for Agentic AI
2026-09-16
PentestChain: A Cost-Aware, MCP-Orchestrated Framework for Automated Penetration Testing with Free-Tier LLMs

Frequent co-authors

Sudha Priyadarshini×1 Rushabh Vipulkumar Patel×1 Dipo Dunsin×1 Mohammed Almaiah×1 Mohamed Chahine Ghanem×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 scores derive from a sparse bibliometric snapshot covering only 2 catalogued 2026 arXiv papers; no peer-reviewed venue record is captured.
Output, impact, and activity dimensions lack citation- and volume-level data and are left unscored rather than estimated.
Institution, country, and career history are unavailable; the profile is name-resolved from public bibliographic data via ORCID 0000-0002-7067-7848.
Collaboration metrics rest on 4 distinct coauthors; the frequent-coauthor list may contain name-matching noise and should be read as indicative.