tile-ascad-fk

Dataset Description

ASCAD v1 Full (raw traces) converted to a unified TileDB Group. Both traces and metadata are stored as TileDB arrays within the same logical group.

Dataset Details

• Format: TileDB Group
• Structure:
  • traces: 2D Dense Array (float/int)
  • metadata: 1D Array (plaintext, key, etc. stored as binary attributes)
• Number of traces: 60,000
• Trace length: 100,000 samples per trace

Files

The dataset is uploaded as a folder structure representing the TileDB Group:

• traces/
• metadata/
• __tiledb_group.tdb

Usage

Use tiledb to open the group or individual arrays.

import tiledb
import numpy as np

# Download/Path to dataset folder
group_uri = "."  # or path to downloaded folder

# 1. Open the Group
with tiledb.Group(group_uri, mode='r') as G:
    print("Group members:", [m.name for m in G])

# 2. Read Traces (Lazy slicing)
with tiledb.DenseArray(f"{group_uri}/traces", mode='r') as A:
    # Read first 10 traces
    traces = A[0:10, :]
    print("Traces shape:", traces["trace"].shape)

# 3. Read Metadata
with tiledb.open(f"{group_uri}/metadata", mode='r') as meta:
    # Read metadata for first 5 traces
    # Returns a dictionary of numpy arrays (bytes)
    data = meta[0:5] 
    
    # Convert bytes back to numpy arrays if needed
    plaintext_0 = np.frombuffer(data["plaintext"][0], dtype=np.uint8)
    print("Plaintext #0:", plaintext_0)

@dataset{ascad-dlsca-tiledb,
  title={ASCAD v1 Full (Unified TileDB)},
  author={DLSCA Platform},
  year={2026},
  publisher={HuggingFace},
  url={[https://huggingface.co/datasets/](https://huggingface.co/datasets/){repo_id}}
}