FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3145735065> ?p ?o ?g. }

• W3145735065 abstract "Outsourced deep neural networks have been demonstrated to suffer from patch-based trojan attacks, in which an adversary poisons the training sets to inject a backdoor in the obtained model so that regular inputs can be still labeled correctly while those carrying a specific trigger are falsely given a target label. Due to the severity of such attacks, many backdoor detection and containment systems have recently, been proposed for deep neural networks. One major category among them are various model inspection schemes, which hope to detect backdoors before deploying models from non-trusted third-parties. In this paper, we show that such state-of-the-art schemes can be defeated by a so-called Scapegoat Backdoor Attack, which introduces a benign scapegoat trigger in data poisoning to prevent the defender from reversing the real abnormal trigger. In addition, it confines the values of network parameters within the same variances of those from clean model during training, which further significantly enhances the difficulty of the defender to learn the differences between legal and illegal models through machine-learning approaches. Our experiments on 3 popular datasets show that it can escape detection by all five state-of-the-art model inspection schemes. Moreover, this attack brings almost no side-effects on the attack effectiveness and guarantees the universal feature of the trigger compared with original patch-based trojan attacks." @default.
• W3145735065 created "2021-04-13" @default.
• W3145735065 creator A5060268538 @default.
• W3145735065 creator A5060284184 @default.
• W3145735065 creator A5070954480 @default.
• W3145735065 creator A5076497877 @default.
• W3145735065 creator A5080640408 @default.
• W3145735065 creator A5083813347 @default.
• W3145735065 date "2021-04-02" @default.
• W3145735065 modified "2023-09-26" @default.
• W3145735065 title "SGBA: A Stealthy Scapegoat Backdoor Attack against Deep Neural Networks" @default.
• W3145735065 cites W2067713319 @default.
• W3145735065 cites W2748789698 @default.
• W3145735065 cites W2753783305 @default.
• W3145735065 cites W2774423163 @default.
• W3145735065 cites W2807363941 @default.
• W3145735065 cites W2900018096 @default.
• W3145735065 cites W2902351501 @default.
• W3145735065 cites W2934843808 @default.
• W3145735065 cites W2963857521 @default.
• W3145735065 cites W2964041528 @default.
• W3145735065 cites W2966187620 @default.
• W3145735065 cites W2966689772 @default.
• W3145735065 cites W2971661634 @default.
• W3145735065 cites W2979693078 @default.
• W3145735065 cites W2982302101 @default.
• W3145735065 cites W2986013765 @default.
• W3145735065 cites W2990270730 @default.
• W3145735065 cites W2995240473 @default.
• W3145735065 cites W3034258347 @default.
• W3145735065 cites W3034414373 @default.
• W3145735065 cites W3034579202 @default.
• W3145735065 cites W3037830434 @default.
• W3145735065 cites W3084847664 @default.
• W3145735065 cites W3091829162 @default.
• W3145735065 cites W3106646114 @default.
• W3145735065 cites W3111943226 @default.
• W3145735065 cites W3115279145 @default.
• W3145735065 cites W3116854761 @default.
• W3145735065 cites W3118608800 @default.
• W3145735065 cites W3046418752 @default.
• W3145735065 doi "https://doi.org/10.48550/arxiv.2104.01026" @default.
• W3145735065 hasPublicationYear "2021" @default.
• W3145735065 type Work @default.
• W3145735065 sameAs 3145735065 @default.
• W3145735065 citedByCount "0" @default.
• W3145735065 crossrefType "posted-content" @default.
• W3145735065 hasAuthorship W3145735065A5060268538 @default.
• W3145735065 hasAuthorship W3145735065A5060284184 @default.
• W3145735065 hasAuthorship W3145735065A5070954480 @default.
• W3145735065 hasAuthorship W3145735065A5076497877 @default.
• W3145735065 hasAuthorship W3145735065A5080640408 @default.
• W3145735065 hasAuthorship W3145735065A5083813347 @default.
• W3145735065 hasBestOaLocation W31457350651 @default.
• W3145735065 hasConcept C154945302 @default.
• W3145735065 hasConcept C174333608 @default.
• W3145735065 hasConcept C17744445 @default.
• W3145735065 hasConcept C199539241 @default.
• W3145735065 hasConcept C2778498553 @default.
• W3145735065 hasConcept C2781045450 @default.
• W3145735065 hasConcept C2984842247 @default.
• W3145735065 hasConcept C38652104 @default.
• W3145735065 hasConcept C41008148 @default.
• W3145735065 hasConcept C41065033 @default.
• W3145735065 hasConcept C50644808 @default.
• W3145735065 hasConceptScore W3145735065C154945302 @default.
• W3145735065 hasConceptScore W3145735065C174333608 @default.
• W3145735065 hasConceptScore W3145735065C17744445 @default.
• W3145735065 hasConceptScore W3145735065C199539241 @default.
• W3145735065 hasConceptScore W3145735065C2778498553 @default.
• W3145735065 hasConceptScore W3145735065C2781045450 @default.
• W3145735065 hasConceptScore W3145735065C2984842247 @default.
• W3145735065 hasConceptScore W3145735065C38652104 @default.
• W3145735065 hasConceptScore W3145735065C41008148 @default.
• W3145735065 hasConceptScore W3145735065C41065033 @default.
• W3145735065 hasConceptScore W3145735065C50644808 @default.
• W3145735065 hasLocation W31457350651 @default.
• W3145735065 hasLocation W31457350652 @default.
• W3145735065 hasOpenAccess W3145735065 @default.
• W3145735065 hasPrimaryLocation W31457350651 @default.
• W3145735065 hasRelatedWork W1494208182 @default.
• W3145735065 hasRelatedWork W2969023901 @default.
• W3145735065 hasRelatedWork W2970233010 @default.
• W3145735065 hasRelatedWork W2971661634 @default.
• W3145735065 hasRelatedWork W3109235236 @default.
• W3145735065 hasRelatedWork W3145735065 @default.
• W3145735065 hasRelatedWork W3198890194 @default.
• W3145735065 hasRelatedWork W4221166349 @default.
• W3145735065 hasRelatedWork W4226394973 @default.
• W3145735065 hasRelatedWork W3046418752 @default.
• W3145735065 isParatext "false" @default.
• W3145735065 isRetracted "false" @default.
• W3145735065 magId "3145735065" @default.
• W3145735065 workType "article" @default.