Skip to content
View mmohamedkhaled's full-sized avatar

Highlights

  • Pro

Block or report mmohamedkhaled

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. 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
mmohamedkhaled/README.md

Mohamed Khaled

Computer Engineering student at The American University in Cairo · Hardware security research

I work on hardware security, mainly FPGA configuration security and Physically Unclonable Functions (PUFs). My current research is a secure-boot scheme for FPGAs that generates keys from PUFs at boot, so the device needs no dedicated secure key storage. I also study how machine learning breaks hardware security primitives, and I am applying to PhD programs for Fall 2027.


Research interests

  • FPGA configuration security: roots of trust for bitstream loading that can be verified openly.
  • PUFs and hardware identity: designs that stay unpredictable against machine-learning modeling attacks within a practical area and reliability budget.
  • Machine learning and hardware security: using learning-based methods both to attack hardware primitives and to verify hardware designs.

Research experience

The American University in Cairo — Research Assistant (Jun 2025 – present)

  • Developing an open-source programming management unit (PMU) that controls FPGA bitstream configuration, as an open alternative to the proprietary units in Intel and Xilinx devices.
  • Developed a PUF-based key-generation approach for secure boot that removes the need for dedicated secure non-volatile key storage. Paper in preparation for IEEE HOST 2027.
  • Designed, simulated, and benchmarked PUF candidates for the design, comparing area overhead, reliability, and uniqueness.

Microsoft — Undergraduate thesis mentorship (Feb 2026 – present)

  • Improving temporal reasoning in small language models through fine-tuning and retrieval-augmented generation, with Microsoft mentors.

Selected projects

Hardware security

Arbiter PUF Modeling Attack Simulation framework for Arbiter PUFs and machine-learning modeling attacks against them. Using the standard parity (phi) feature transform, a linear model predicts responses of an unprotected Arbiter PUF with 99.3% accuracy; adding fixed stage permutations breaks linear separability and drops the same attack to 52.5%, close to random guessing.

NeuralTram: systolic array for edge AI (private) Parameterized weight-stationary systolic array of INT8 multiply-accumulate cells, targeting SkyWater 130 nm at under 1 mm² and 100 MHz. Written in Yosys-compatible Verilog-2001 and verified with cocotb against a NumPy reference model.

RISC-V RV32I core (co-developer) RV32I processor in Verilog: ALU, control unit, register file, and analysis of pipeline hazards.

Machine learning security

LLM Language Jailbreak Study Defensive measurement study of whether encodings (an invented language, ROT13/Base64, rare languages) bypass refusal training in 5 instruction-tuned models on AdvBench. Uses decode-fidelity checks, multiple judges with inter-rater agreement (kappa), and human adjudication. Harmful completions and model weights are withheld from the public release.

Systems and cryptography

  • cksumx: C++17 checksum tool and library on OpenSSL (SHA-2/3, BLAKE2) with constant-memory streaming and constant-time digest comparison.
  • Hill Cipher: C++17 implementation with a known-plaintext key-recovery attack.
  • Cybersecurity 101: 36 small, tested Python implementations of cryptography, hashing, and number theory.
  • RansomVault-Lab: teaching lab for hybrid encryption (AES-256-GCM with RSA-4096-OAEP key wrapping). It encrypts only local test files and contains no spreading or extortion logic.
  • Attack Graph Analysis: network attack graphs in NetworkX with path weights derived from CVSS scores.
  • LPM-Rust: Linux process manager in Rust with terminal, graphical, and command-line front ends.
  • RescueMesh (course project, private): three-node fault-tolerant system in Rust with Raft-style leader election, compared against Bully and Chang–Roberts election; 100 automated tests with fault injection.
  • Logic Circuits Simulator: event-driven gate-level simulator with propagation delays.

Industry experience

Groupe SEB — Information Security Intern (Jun – Sep 2026)

  • Performed an authorized black/gray-box audit of a production multi-tenant e-commerce platform over five rounds: 45 findings (11 Critical, 13 High) scored with CVSS v3.1, including three end-to-end attack chains.
  • Extended an in-house Python penetration-testing toolkit (249 scanner modules across web/API, cloud, and Active Directory) and set up on-premises LLM hosting so findings could be triaged without sending data to external APIs.

Xanadu Group — Full Stack Engineer Intern (Jun – Sep 2026)

  • Built the corporate website in Next.js with a Node.js backend protected by rate limiting, CSRF protection, and input sanitization.

Eden MEA — AI Engineering Intern (Jan – Feb 2026)

  • Built a financial data pipeline (n8n, AWS RDS) and an orchestration layer for AI agents.

Education

The American University in Cairo — B.Sc. Computer Engineering, minors in Mathematics and Business (2022 – 2027) GPA 3.77/4.00 · USAID Merit Award Relevant courses: Security in Computing, AI Security, Computer Architecture, Digital Design, Computer Organization & Assembly, Operating Systems, Distributed Systems, Artificial Intelligence

University of Central Florida — Exchange semester, Computer Engineering (Spring 2025)


Skills

Area Tools
Hardware Verilog, FPGA, Vivado, Yosys, OpenFPGA, Icarus Verilog, cocotb, SkyWater 130 nm
Hardware security PUFs, fuzzy extractors (BCH), secure bitstream loading (AES-256, SHA-256, HMAC), JTAG
Security Penetration testing, LLM red-teaming, OWASP Top 10, CVSS v3.1
Machine learning PyTorch, Hugging Face, LoRA fine-tuning, RAG, NumPy, Pandas
Languages Python, C++, C, Rust, Verilog, SQL, TypeScript

Leadership

Enactus AUC — President (Jun 2025 – present), previously Project Director for two years. Capsim Business Challenge — global top 3 (May 2024).


Contact

Email: mohamedabdelfatah572@aucegypt.edu · LinkedIn

Pinned Loading

  1. RansomVault-Lab RansomVault-Lab Public

    Educational Python toolkit for studying modern file encryption and the ransomware crypto model : AES-256-GCM + RSA-4096 hybrid key wrapping. Safe by design

    Python

  2. arbiter-puf-modeling-attack arbiter-puf-modeling-attack Public

    How well do XOR-masking, noise, and stage permutation stop the classic LR attack on an Arbiter PUF? A reproducible 30-trial benchmark.

    Jupyter Notebook

  3. GM-Sniper/RISCV32I-CPU GM-Sniper/RISCV32I-CPU Public

    Implementation of a RISC-V CPU on Verilog using Xilinx Artix-7 FPGA board

    Verilog 5 1