FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2948811271> ?p ?o ?g. }

• W2948811271 abstract "Deep neural networks (DNNs) have been shown to tolerate brain damage: cumulative changes to the network's parameters (e.g., pruning, numerical perturbations) typically result in a graceful degradation of classification accuracy. However, the limits of this natural resilience are not well understood in the presence of small adversarial changes to the DNN parameters' underlying memory representation, such as bit-flips that may be induced by hardware fault attacks. We study the effects of bitwise corruptions on 19 DNN models---six architectures on three image classification tasks---and we show that most models have at least one parameter that, after a specific bit-flip in their bitwise representation, causes an accuracy loss of over 90%. We employ simple heuristics to efficiently identify the parameters likely to be vulnerable. We estimate that 40-50% of the parameters in a model might lead to an accuracy drop greater than 10% when individually subjected to such single-bit perturbations. To demonstrate how an adversary could take advantage of this vulnerability, we study the impact of an exemplary hardware fault attack, Rowhammer, on DNNs. Specifically, we show that a Rowhammer enabled attacker co-located in the same physical machine can inflict significant accuracy drops (up to 99%) even with single bit-flip corruptions and no knowledge of the model. Our results expose the limits of DNNs' resilience against parameter perturbations induced by real-world fault attacks. We conclude by discussing possible mitigations and future research directions towards fault attack-resilient DNNs." @default.
• W2948811271 created "2019-06-14" @default.
• W2948811271 creator A5007374820 @default.
• W2948811271 creator A5033409139 @default.
• W2948811271 creator A5037829512 @default.
• W2948811271 creator A5075711534 @default.
• W2948811271 creator A5083941826 @default.
• W2948811271 date "2019-06-03" @default.
• W2948811271 modified "2023-10-18" @default.
• W2948811271 title "Terminal Brain Damage: Exposing the Graceless Degradation in Deep Neural Networks Under Hardware Fault Attacks" @default.
• W2948811271 cites W1563795667 @default.
• W2948811271 cites W1605005685 @default.
• W2948811271 cites W1677182931 @default.
• W2948811271 cites W1836465849 @default.
• W2948811271 cites W1921523184 @default.
• W2948811271 cites W1946342668 @default.
• W2948811271 cites W2024109642 @default.
• W2948811271 cites W2067713319 @default.
• W2948811271 cites W2095705004 @default.
• W2948811271 cites W2112507308 @default.
• W2948811271 cites W2112796928 @default.
• W2948811271 cites W2114766824 @default.
• W2948811271 cites W2117539524 @default.
• W2948811271 cites W2119112357 @default.
• W2948811271 cites W2124136621 @default.
• W2948811271 cites W2157116240 @default.
• W2948811271 cites W2162552722 @default.
• W2948811271 cites W2163563130 @default.
• W2948811271 cites W2183341477 @default.
• W2948811271 cites W2194775991 @default.
• W2948811271 cites W2276892413 @default.
• W2948811271 cites W2300242332 @default.
• W2948811271 cites W2329308213 @default.
• W2948811271 cites W2491829854 @default.
• W2948811271 cites W2505343551 @default.
• W2948811271 cites W2515385951 @default.
• W2948811271 cites W2516668814 @default.
• W2948811271 cites W2525778437 @default.
• W2948811271 cites W2533598788 @default.
• W2948811271 cites W2537014044 @default.
• W2948811271 cites W2750990141 @default.
• W2948811271 cites W2751092112 @default.
• W2948811271 cites W2767260595 @default.
• W2948811271 cites W2771112233 @default.
• W2948811271 cites W2790483052 @default.
• W2948811271 cites W2795222486 @default.
• W2948811271 cites W2801117814 @default.
• W2948811271 cites W2801268510 @default.
• W2948811271 cites W2807363941 @default.
• W2948811271 cites W2807765471 @default.
• W2948811271 cites W2807835252 @default.
• W2948811271 cites W2809188712 @default.
• W2948811271 cites W2886541648 @default.
• W2948811271 cites W2888940765 @default.
• W2948811271 cites W2889797931 @default.
• W2948811271 cites W2889929196 @default.
• W2948811271 cites W2899469948 @default.
• W2948811271 cites W2962726564 @default.
• W2948811271 cites W2962835968 @default.
• W2948811271 cites W2963110888 @default.
• W2948811271 cites W2963343288 @default.
• W2948811271 cites W2963709416 @default.
• W2948811271 cites W2963771448 @default.
• W2948811271 cites W2963777610 @default.
• W2948811271 cites W2963888996 @default.
• W2948811271 cites W2964228333 @default.
• W2948811271 cites W2964299589 @default.
• W2948811271 cites W2964318098 @default.
• W2948811271 cites W3118608800 @default.
• W2948811271 hasPublicationYear "2019" @default.
• W2948811271 type Work @default.
• W2948811271 sameAs 2948811271 @default.
• W2948811271 citedByCount "8" @default.
• W2948811271 countsByYear W29488112712020 @default.
• W2948811271 countsByYear W29488112712021 @default.
• W2948811271 crossrefType "posted-content" @default.
• W2948811271 hasAuthorship W2948811271A5007374820 @default.
• W2948811271 hasAuthorship W2948811271A5033409139 @default.
• W2948811271 hasAuthorship W2948811271A5037829512 @default.
• W2948811271 hasAuthorship W2948811271A5075711534 @default.
• W2948811271 hasAuthorship W2948811271A5083941826 @default.
• W2948811271 hasConcept C108010975 @default.
• W2948811271 hasConcept C108583219 @default.
• W2948811271 hasConcept C111919701 @default.
• W2948811271 hasConcept C113775141 @default.
• W2948811271 hasConcept C11413529 @default.
• W2948811271 hasConcept C121332964 @default.
• W2948811271 hasConcept C127313418 @default.
• W2948811271 hasConcept C127705205 @default.
• W2948811271 hasConcept C134765980 @default.
• W2948811271 hasConcept C154945302 @default.
• W2948811271 hasConcept C165205528 @default.
• W2948811271 hasConcept C175551986 @default.
• W2948811271 hasConcept C17744445 @default.
• W2948811271 hasConcept C199360897 @default.
• W2948811271 hasConcept C199539241 @default.
• W2948811271 hasConcept C2776359362 @default.
• W2948811271 hasConcept C2779585090 @default.
• W2948811271 hasConcept C2779679103 @default.
• W2948811271 hasConcept C2984842247 @default.

• next