Researchers propose SSTA to deceive cognitive radars across architectures
Original title: Bridging Digital Deception and Physical Fidelity: A Spatio-Temporal Attack on Cognitive Radars
Authors report a 10.36% higher average black-box success rate and 0 dB main-lobe attenuation after matched filtering.
A cs.CR paper presents Segmented Spatio-Temporal Attack (SSTA), aimed at deep networks that process cognitive-radar baseband sequences. The method splits global and local features and folds channel distortion into differentiable operators using Expectation over Transformation. Across nine CNN, RNN and Transformer models, the authors say SSTA beat their state-of-the-art baseline by 10.36% in average black-box attack success rate. Baseband simulations reportedly kept main-lobe attenuation at 0 dB and noise within CFAR limits.
Unconfirmed: Success figures come from the authors’ cross-evaluations and baseband simulations, not confirmed operational use.
- Evaluations covered nine models spanning CNNs, RNNs and Transformers.
- Reported gain over the baseline was 10.36% average black-box attack success rate.
- Simulated waveforms showed 0 dB main-lobe attenuation after matched filtering.
- Noise-floor rise was described as staying inside CFAR tolerances.
From the original article
1 min read at arxiv.orgAs cognitive radars evolve towards the end-to-end processing of one-dimensional complex baseband sequences, deep networks face severe adversarial threats. However, existing attacks encounter dual limitations in practical electronic warfare: low cross-architecture transferability and severe vulnerability to physical channel distortions, which degrades their tactical reliability.
By Tongrun Xing · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.