Three defenses harden face recognition
Original title: A Bag of Tricks for Defending Deep Learning-based Face Recognition Pipelines against Backdoor Attacks
Three tricks detect poisoned identities at training and filter inputs at inference.
Backdoor attacks via data poisoning are targeted. The first defense detects poisoned identities during training. The last two filter suspicious inputs at inference before responses. Each method includes motivation, algorithms and evaluation on realistic face recognition workloads.
- Training-time detection of poisoned identities
- Inference-time input filtering
- Targets third-party training data
- Practical hardening for providers
- Evaluated on face recognition pipelines
From the original article
1 min read at arxiv.orgThe protection of mature, multi-stage Face Recognition Systems is crucial to system providers. This technical report targets the threat model most relevant to industry practitioners: although model training is performed in-house, training data is sourced from possibly-compromised third parties. Given this setting, we present three practical defense "tricks" against data poisoning-based backdoor attacks.
By Quentin Le Roux · 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.