Skip to main content

ICARUS Interference Detection (INTD) & Classification (INTC)

Interference detection and classifcation benchmark from the ICARUS dataset (IQ + cycle frequency features).

  • Raw data: https://genesys-lab.org/ICARUS
  • Paper: INFOCOM 2023 — ICARUS: Learning on IQ and Cycle Frequencies for Detecting Anomalous RF Underlay Signals

Tasks supported​

  • Interference detection / Interference Classifcation

Class labels​

  • 0: LTE only (no interference)
  • 1: DSSS interference (type 1; log2(DSSS_Mod)=1)
  • 2: DSSS interference (type 2; log2(DSSS_Mod)=2)

Split​

  • 80/20 random split

Preprocessing​

  1. Scan batch folders for matching Metadata/*.csv and IQ/*.bin files.
  2. Infer IQ dtype from file size, then load interleaved I/Q samples.
  3. Center-crop each sample to a fixed length (default 4096).
  4. Normalize I and Q using dataset-wide mean/std constants.
  5. Encode labels: LTE-only = 0; DSSS interference uses log2(DSSS_Mod).
  6. Save samples, labels, band (5/10 MHz), and sampling rate to HDF5.

Script: preprocess_icarus.py

python preprocessing/preprocess_icarus.py \
  --data-path <ICARUS_ROOT> \
  --output data/icarus.h5 \
  --max-len 4096

Metric​

  • Interference detection accuracy (whether interference exits) on test split.
  • Interference classifcation accuracy (type of interference) on test split.

Citation​

@INPROCEEDINGS{10228929,
  author={Roy, Debashri and Chaudhury, Vini and Tassie, Chinenye and Spooner, Chad and Chowdhury, Kaushik},
  booktitle={IEEE INFOCOM 2023 - IEEE Conference on Computer Communications}, 
  title={ICARUS: Learning on IQ and Cycle Frequencies for Detecting Anomalous RF Underlay Signals}, 
  year={2023},
  pages={1-10},
  doi={10.1109/INFOCOM53939.2023.10228929}
}