Memristive Security Framework

A research-grade simulation suite for fault-aware memristive crossbars, side-channel attacks, and PUF-based hardware security countermeasures, built on MemTorch + PyTorch.

Highlights

• MemTorch-based mapping of DNN weights onto resistive-RAM crossbars
• Stuck-at (SA0/SA1) fault injection with configurable fault rates
• Correlation Power Analysis (CPA) on simulated crossbar current traces
• Memristive PUF generation, CRP database, and metric evaluation
• Key result: characterised the Fault Rate vs. Security (CPA success / PUF reliability) trade-off

Setup

conda env create -f environment.yml
conda activate memristive-security

Running the orchestrator

python src/main.py --config config/default.yaml

Running tests

pytest tests/ -v

Notebooks

Notebook	Topic
01_memtorch_baseline.ipynb	MemTorch mapping & inference accuracy
02_fault_injection_analysis.ipynb	Stuck-at fault sweep
03_side_channel_attack.ipynb	CPA demo
04_fault_security_cross.ipynb	Fault–Security trade-off
05_puf_evaluation.ipynb	PUF uniqueness & reliability

Project Structure

memristive-security-framework/
├── config/          # YAML hyper-parameters
├── src/             # Python source package
│   ├── models/      # Neural-network & mapping utilities
│   ├── hardware/    # Crossbar, fault, variability
│   ├── attacks/     # CPA & leakage models
│   ├── defense/     # PUF generation & metrics
│   └── utils/       # Data, metrics, plotting
├── notebooks/
├── tests/
├── scripts/
├── results/
└── docs/