Stores `tracks/track_0/data/raw_data` with shuffle + gzip (level 4) instead of Blosc in the four `data/Fundamental/` files that the Hub's ClamAV scan marks unsafe (`Win.Exploit.CVE_2016_4332_1-6028013-0`):
- `126987_3_Focused_Uncoded_TX.hdf5`
- `559277_2_Focused_Uncoded_TX.hdf5`
- `729304_2_Focused_Uncoded_TX.hdf5`
- `829312_1_Focused_Uncoded_TX.hdf5`
The matching bytes sit inside Blosc-compressed `raw_data` chunks, not in the HDF5 object headers the CVE is about, so this is a false positive (reported to ClamAV). Re-encoding with Blosc at another level or shuffle mode only clears some files (#82). gzip changes the chunk bytes completely and is the same fix as #83 for UBC `session_01_f0374.hdf5`.
Files changed: 4. Only the compression filter of `raw_data` changed. Its values, dtype, shape, chunk shape and fill value, every other dataset (still Blosc) and every attribute are untouched. The Sept 26 subject-ID fix (#74) is kept, since the files were rewritten from current `main`.
Check…
21 days ago
vanderbilt: subject/id is the six-digit patient ID, without the view number (#74)
- vanderbilt: subject/id is the six-digit patient ID, without the view number (d82df6cea24a686a59a59d46b74c79078c3b45bd)
Co-authored-by: Oisín Nolan <oisinn@users.noreply.huggingface.co>
25 days ago
vanderbilt: add anatomy annotation, and mark the CIRS scan as a phantom (#17)
- vanderbilt: add anatomy annotation, and mark the CIRS scan as a phantom (775816c1a94cd5468d424b0d4552c81d01fa14c5)
Co-authored-by: Oisín Nolan <oisinn@users.noreply.huggingface.co>
