Skip to content
View 61465's full-sized avatar
🏠
Working from home
🏠
Working from home

Block or report 61465

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
61465/README.md
GZ

Abd Alrahaman Mohamed

Full-Stack Engineer · AI Systems · Applied Security Research

Alexandria, Egypt · Remote-first


Impact

  • 🏆 Formalized APDS — Asymmetric Patch Detection for Symmetric APIs (new vulnerability-discovery methodology). First application: an un-patched symmetric sibling of CVE-2024-40658, byte-identical across 6 AOSP branches for 25 months. Submitted to Google ASR (CVSS 10.0 Critical · in panel review).
  • 🛠️ Open-sourced the reference implementation at github.com/61465/apds-hunt within 24 h of submission (90-day SLA).
  • ✅ PostHog PR #5083 merged (opt-in WebView bot detection, 11 days end-to-end).
  • 🚀 PostHog PR #5244 open — ported upstream rrweb #1652 to fix a 9–10 s session-replay freeze on 13 k-node mounts (MutationBuffer.processBufferedMutations O(n²) → O(n)).
  • 🚀 Supabase storage PR #1488 open — auto-select URL-signing JWK from JWT_JWKS (closes #629, open 607 days). 25 new tests including a JOSE round-trip proof.
  • 🚀 Supabase storage PR #1439 open — UTF-8 object-key hardening + NFC normalization + lone-surrogate rejection.
  • 📝 Trigger.dev GHSA-65rr-73jq-6qhf accepted (CVSS 9.6). Documenso GHSA-2rp7-mrh6-v9m2 in triage (CVSS 7.6, first-ever GHSA on the repo).

Impact

  • Top 16% (325 / 2,076 teams) on Kaggle AI Agent Security: Multi-Step Tool Attacks — verified score 83.745
  • Reduced WhatsApp ban rate from Meta's 0.5% baseline to 0.1% (5× improvement) via a 13-layer defense stack in production SaaS
  • 90.6% adversarial containment at 5.9% FPR on 8,000 seeded sessions — Ariadne framework (SSRN preprint)
  • Hospital-ready radiology AI: 69 s per 24-slice CT at 92% confidence — midcine v3
  • Shipped 134-agent NEXUS catalog with MCP orchestration + a five-verifier grounding gate that removed 724 unsupported cyber claims across 770 production invocations

Research


Stack

Languages · Python · TypeScript · JavaScript · Rust · Dart · SQL · LaTeX AI / ML · PyTorch · Transformers · LoRA · DPO · Ollama · Claude API · MCP · RAG · LangGraph · multi-agent orchestration Web / Backend · Next.js 15 · React · NestJS · FastAPI · Node.js · Express · Prisma · PWA · Turborepo · pnpm workspaces Healthcare · DICOM · pynetdicom · highdicom · OHIF v3 · Cornerstone3D · FHIR R4 · HL7 Cybersecurity · WebAuthn/Passkeys · OAuth2 · SIEM · SOAR · EDR · YARA · pen testing (TCM-certified) · adversarial-AI defense · responsible disclosure (CVSS 3.1) DevOps · Docker · Kubernetes · Firecracker · gVisor · Tailscale mesh · Cloudflare Tunnel · self-hosted PaaS (Ollama + Vaultwarden + Gitea + Grafana)


Elsewhere

LinkedIn Kaggle SSRN Email


LiveProducts


Pinned

Selected showcases from the ecosystem. Full source code, datasets, and proprietary models remain private for commercial and research reasons.

Popular repositories Loading

  1. lucid lucid Public

    Python 1

  2. read read Public

    HTML 1

  3. game-zone-hub game-zone-hub Public

    HTML 1

  4. projects projects Public

    HTML 1

  5. Ariadne Ariadne Public

    Adaptive Human-in-the-Loop Defense for AI Agents (SSRN 6870178)

    1

  6. midcine midcine Public

    Arabic Cloud-Native RIS/PACS — Radiology AI Platform (showcase)

    HTML 1