BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain¶
Gu, T., Dolan-Gavitt, Brendan, & Garg, S. (2017). BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain.
Cited by¶
3 citations across 3 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Trusted Intermediary Compromise
- AI supply chain. Poisoned training data ingested through scraped corpora, trojaned model checkpoints used without artefact-level verification, and benchmark contamination that produces falsely trustworthy numbers.
This sourceDemonstrates trojaned model checkpoints and poisoned training producing hidden backdoors in the ML supply chain.
- AI supply chain. Poisoned training data ingested through scraped corpora, trojaned model checkpoints used without artefact-level verification, and benchmark contamination that produces falsely trustworthy numbers.
Domain-specific¶
- AI Supply-Chain Attack
- The BadNets result (Gu, Dolan-Gavitt, and Garg, 2017) is the seminal demonstration on the learned-behaviour surface
This sourcePresents BadNets, a maliciously trained network that performs as well as a clean one on ordinary inputs but behaves badly on attacker-chosen inputs, demonstrated on a street-sign classifier. Demonstrates a training-set-poisoned street-sign classifier in which a small yellow sticker on a stop sign makes the stop sign read as a speed-limit sign.
Supported in partVerified against a saved copy of the source
“BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain”
- The BadNets result (Gu, Dolan-Gavitt, and Garg, 2017) is the seminal demonstration on the learned-behaviour surface
- Data Poisoning Attack
- But any stop sign bearing the sticker trigger at inference time was confidently misread as, for instance, a speed-limit sign
This sourceThe BadNets traffic-sign backdoor, in which more than 90% of stop signs bearing a Post-it-sized sticker trigger were classified as speed-limit signs, one real sign with 95% confidence.
Supported in partVerified against the source
- But any stop sign bearing the sticker trigger at inference time was confidently misread as, for instance, a speed-limit sign
Verification¶
Does it exist? Not checked yet. This entry carries no identifier to resolve. It was extracted from the citation as written in the article, normalized, and deduplicated against the rest of the registry.
Does it back the claim? Read against the text for 2 of 3 citations: 2 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:bea631042d7e · see in the full table