{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,9]],"date-time":"2026-04-09T00:01:43Z","timestamp":1775692903052,"version":"3.50.1"},"reference-count":45,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2024,11,15]],"date-time":"2024-11-15T00:00:00Z","timestamp":1731628800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2024,11,15]],"date-time":"2024-11-15T00:00:00Z","timestamp":1731628800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"name":"BECA FPI PROGRAMA PROPIO DE LA UNIVERSITAT POLIT\u00c8CNICA DE VAL\u00c8NCIA \u2013 SUBPROGRAMA 1","award":["20210037"],"award-info":[{"award-number":["20210037"]}]},{"name":"Valencian government project \"IFAC\": Implementing Fault-Tolerant Autonomous Computers","award":["CISEJI\/2022\/30"],"award-info":[{"award-number":["CISEJI\/2022\/30"]}]},{"name":"Spanish Ministry of Science, Innovation and Universities' \u201cRam\u00f3n y Cajal\u201d fellowship","award":["RYC2020-030685-I"],"award-info":[{"award-number":["RYC2020-030685-I"]}]},{"DOI":"10.13039\/501100004233","name":"Universitat Polit\u00e8cnica de Val\u00e8ncia","doi-asserted-by":"crossref","id":[{"id":"10.13039\/501100004233","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["J Supercomput"],"published-print":{"date-parts":[[2025,1]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:p>Neural networks are widely used in critical environments such as healthcare, autonomous vehicles, or video surveillance. To ensure the safety of the systems that rely on their functionality, it is essential to validate their correct behaviour in the presence of faults. This paper studies the behaviour of state-of-the-art neural network models with fault injection in their weights. For this purpose, we analyse the sensitivity of these models and identify the impact of bit flips on their accuracy. To mitigate the effects of faults, we introduce two mechanisms that leverage bit-level redundancy for protection. The first mechanism, <jats:bold>Fixed Protection<\/jats:bold>, safeguards consecutive sets of bits, while the second, <jats:bold>Variable Protection<\/jats:bold>, targets non-consecutive bits. Our findings demonstrate that, on average, random bit flip faults cause the accuracy of the original models to drop by 1.3% to over 3%. However, with our protection mechanisms in place, accuracy reductions are significantly minimised, ranging from only 0.0001% to 0.4%.<\/jats:p>","DOI":"10.1007\/s11227-024-06693-7","type":"journal-article","created":{"date-parts":[[2024,11,15]],"date-time":"2024-11-15T11:34:39Z","timestamp":1731670479000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":6,"title":["Exploiting neural networks bit-level redundancy to mitigate the impact of faults at inference"],"prefix":"10.1007","volume":"81","author":[{"given":"Izan","family":"Catal\u00e1n","sequence":"first","affiliation":[]},{"given":"Jos\u00e9","family":"Flich","sequence":"additional","affiliation":[]},{"given":"Carles","family":"Hern\u00e1ndez","sequence":"additional","affiliation":[]}],"member":"297","published-online":{"date-parts":[[2024,11,15]]},"reference":[{"key":"6693_CR1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2016.2577031","author":"S Ren","year":"2017","unstructured":"Ren S, He K, Girshick R, Sun J (2017) Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Trans Pattern Anal Mach Intelli. https:\/\/doi.org\/10.1109\/TPAMI.2016.2577031","journal-title":"IEEE Trans Pattern Anal Mach Intelli"},{"key":"6693_CR2","doi-asserted-by":"publisher","unstructured":"Vasuki A, Govindaraju S (2017) Deep neural networks for image classification, pp 27\u201349. https:\/\/doi.org\/10.3233\/978-1-61499-822-8-27","DOI":"10.3233\/978-1-61499-822-8-27"},{"key":"6693_CR3","unstructured":"Dosovitskiy A, Beyer L, Kolesnikov A, Weissenborn D, Zhai X, Unterthiner T, Dehghani M, Minderer M, Heigold G, Gelly S, Uszkoreit J, Houlsby N (2021) An image is worth 16x16 words: transformers for image recognition at scale. In: 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3\u20137. OpenReview.net. https:\/\/openreview.net\/forum?id=YicbFdNTTy"},{"key":"6693_CR4","doi-asserted-by":"publisher","unstructured":"Rakin AS, He Z, Fan D (2019) Bit-flip attack: crushing neural network with progressive bit search. In: 2019 IEEE\/CVF International Conference on Computer Vision (ICCV), pp 1211\u20131220. https:\/\/doi.org\/10.1109\/ICCV.2019.00130","DOI":"10.1109\/ICCV.2019.00130"},{"key":"6693_CR5","doi-asserted-by":"publisher","unstructured":"Hong S, Frigo P, Kaya Y, Giuffrida C, Dumitra\u015f T (2019) Terminal brain damage: exposing the graceless degradation in deep neural networks under hardware fault attacks. In: Proceedings of the 28th USENIX Conference on Security Symposium, Santa Clara, CA, USA, pp 497\u2013514. https:\/\/doi.org\/10.5555\/3361338.3361373","DOI":"10.5555\/3361338.3361373"},{"key":"6693_CR6","doi-asserted-by":"publisher","DOI":"10.1145\/3126908.3126964","volume-title":"Understanding error propagation in deep learning neural network (DNN) accelerators and applications","author":"G Li","year":"2017","unstructured":"Li G, Hari SKS, Sullivan M, Tsai T, Pattabiraman K, Emer J, Keckler SW (2017) Understanding error propagation in deep learning neural network (DNN) accelerators and applications. Association for Computing Machinery, New York. https:\/\/doi.org\/10.1145\/3126908.3126964"},{"key":"6693_CR7","doi-asserted-by":"publisher","unstructured":"Hoang L-H, Hanif MA, Shafique M (2020) Ft-clipact: resilience analysis of deep neural networks and improving their fault tolerance using clipped activation. In: Proceedings of the 23rd Conference on Design, Automation and Test in Europe. DATE \u201920, pp 1241\u20131246. EDA Consortium, San Jose, CA, USA. https:\/\/doi.org\/10.23919\/DATE48585.2020.9116571","DOI":"10.23919\/DATE48585.2020.9116571"},{"key":"6693_CR8","doi-asserted-by":"publisher","unstructured":"Salay R, Queiroz R, Czarnecki K (2017) An analysis of ISO 26262: using machine learning safely in automotive software. https:\/\/doi.org\/10.4271\/9780768002683","DOI":"10.4271\/9780768002683"},{"key":"6693_CR9","unstructured":"Stock P (2021) Efficiency and redundancy in deep learning models: theoretical considerations and practical applications. Theses, Universit\u00e9 de Lyon. https:\/\/theses.hal.science\/tel-03208517"},{"key":"6693_CR10","unstructured":"Geissler F, Qutub SS, Roychowdhury S, Khoshouyeh AA, Peng Y, Dhamasia A, Graefe R, Pattabiraman K, Paulitsch M (2021) Towards a safety case for hardware fault tolerance in convolutional neural networks using activation range supervision. CoRR abs\/2108.07019"},{"key":"6693_CR11","doi-asserted-by":"publisher","unstructured":"Li W, Ge G, Guo K, Chen X, Wei Q, Gao Z, Wang Y, Yang H (2020) Soft error mitigation for deep convolution neural network on FPGA accelerators. In: 2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), pp 1\u20135. https:\/\/doi.org\/10.1109\/AICAS48895.2020.9073925","DOI":"10.1109\/AICAS48895.2020.9073925"},{"key":"6693_CR12","doi-asserted-by":"publisher","first-page":"20190164","DOI":"10.1098\/rsta.2019.0164","volume":"378","author":"M Hanif","year":"2019","unstructured":"Hanif M, Shafique M (2019) Salvagednn: salvaging deep neural network accelerators with permanent faults through saliency-driven fault-aware mapping. Philos Trans R Soc A Math Phys Eng Sci 378:20190164. https:\/\/doi.org\/10.1098\/rsta.2019.0164","journal-title":"Philos Trans R Soc A Math Phys Eng Sci"},{"key":"6693_CR13","doi-asserted-by":"publisher","unstructured":"Liu Q, Wen W, Wang Y (2020) Concurrent weight encoding-based detection for bit-flip attack on neural network accelerators. In: 2020 IEEE\/ACM International Conference on Computer Aided Design (ICCAD), pp 1\u20138. https:\/\/doi.org\/10.1145\/3400302.3415726","DOI":"10.1145\/3400302.3415726"},{"key":"6693_CR14","doi-asserted-by":"publisher","unstructured":"Koopman P, Chakravarty T (2004) Cyclic redundancy code (CRC) polynomial selection for embedded networks. In: International Conference on Dependable Systems and Networks, 2004, pp 145\u2013154. https:\/\/doi.org\/10.1109\/DSN.2004.1311885","DOI":"10.1109\/DSN.2004.1311885"},{"key":"6693_CR15","doi-asserted-by":"publisher","DOI":"10.1109\/TNS.2016.2547963","author":"A Neale","year":"2016","unstructured":"Neale A, Sachdev M (2016) Neutron radiation induced soft error rates for an adjacent-ECC protected SRAM in 28 nm CMOS. IEEE Trans Nucl Sci. https:\/\/doi.org\/10.1109\/TNS.2016.2547963","journal-title":"IEEE Trans Nucl Sci"},{"key":"6693_CR16","doi-asserted-by":"publisher","first-page":"77","DOI":"10.1007\/978-3-319-52153-4_5","volume-title":"Feeding two cats with one bowl: on designing a fault and side-channel resistant software encoding scheme","author":"J Breier","year":"2017","unstructured":"Breier J, Hou X (2017) Feeding two cats with one bowl: on designing a fault and side-channel resistant software encoding scheme. Springer, Cham, pp 77\u201394. https:\/\/doi.org\/10.1007\/978-3-319-52153-4_5"},{"issue":"3","key":"6693_CR17","doi-asserted-by":"publisher","first-page":"1065","DOI":"10.1109\/TDSC.2019.2897663","volume":"18","author":"J Breier","year":"2021","unstructured":"Breier J, Hou X, Liu Y (2021) On evaluating fault resilient encoding schemes in software. IEEE Trans Dependable Secur Comput 18(3):1065\u20131079. https:\/\/doi.org\/10.1109\/TDSC.2019.2897663","journal-title":"IEEE Trans Dependable Secur Comput"},{"issue":"10","key":"6693_CR18","doi-asserted-by":"publisher","first-page":"1595","DOI":"10.1109\/TCAD.2013.2263037","volume":"32","author":"X Guo","year":"2013","unstructured":"Guo X, Karri R (2013) Recomputing with permuted operands: a concurrent error detection approach. IEEE Trans Comput-Aided Des Integr Circuits Syst 32(10):1595\u20131608. https:\/\/doi.org\/10.1109\/TCAD.2013.2263037","journal-title":"IEEE Trans Comput-Aided Des Integr Circuits Syst"},{"key":"6693_CR19","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA.2012.6237031","volume-title":"A defect-tolerant accelerator for emerging high-performance applications","author":"O Temam","year":"2012","unstructured":"Temam O (2012) A defect-tolerant accelerator for emerging high-performance applications. Association for Computing Machinery, New York. https:\/\/doi.org\/10.1109\/ISCA.2012.6237031"},{"key":"6693_CR20","doi-asserted-by":"publisher","unstructured":"Kim J-S, Yang J-S (2019) Dris-3: deep neural network reliability improvement scheme in 3d die-stacked memory based on fault analysis, pp 1\u20136. https:\/\/doi.org\/10.1145\/3316781.3317805","DOI":"10.1145\/3316781.3317805"},{"key":"6693_CR21","doi-asserted-by":"publisher","unstructured":"He Z, Rakin AS, Li J, Chakrabarti C, Fan D (2020) Defending and harnessing the bit-flip based adversarial weight attack. In: 2020 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp 14083\u201314091. https:\/\/doi.org\/10.1109\/CVPR42600.2020.01410","DOI":"10.1109\/CVPR42600.2020.01410"},{"key":"6693_CR22","doi-asserted-by":"publisher","unstructured":"Li J, Rakin AS, Xiong Y, Chang L, He Z, Fan D, Chakrabarti C (2020) Defending bit-flip attack through DNN weight reconstruction. In: 2020 57th ACM\/IEEE Design Automation Conference (DAC), pp 1\u20136. https:\/\/doi.org\/10.1109\/DAC18072.2020.9218665","DOI":"10.1109\/DAC18072.2020.9218665"},{"key":"6693_CR23","doi-asserted-by":"publisher","unstructured":"Khare Y, Lakara K, Inukonda MS, Mittal S, Chandra M, Kaushik A (2022) Design and analysis of novel bit-flip attacks and defense strategies for DNNs. In: 2022 IEEE Conference on Dependable and Secure Computing (DSC), pp 1\u20138. https:\/\/doi.org\/10.1109\/DSC54232.2022.9888943","DOI":"10.1109\/DSC54232.2022.9888943"},{"key":"6693_CR24","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2022.3211411","author":"L Liu","year":"2023","unstructured":"Liu L, Guo Y, Cheng Y, Zhang Y, Yang J (2023) Generating robust DNN with resistance to bit-flip based adversarial weight attack. IEEE Trans Comput. https:\/\/doi.org\/10.1109\/TC.2022.3211411","journal-title":"IEEE Trans Comput"},{"key":"6693_CR25","doi-asserted-by":"publisher","unstructured":"Li Y, Li M, Luo B, Tian Y, Xu Q (2020) Deepdyve: dynamic verification for deep neural networks. In: Proceedings of the 2020 ACM SIGSAC Conference on Computer and Communications Security. CCS \u201920, pp 101\u2013112. https:\/\/doi.org\/10.1145\/3372297.3423338","DOI":"10.1145\/3372297.3423338"},{"key":"6693_CR26","doi-asserted-by":"publisher","unstructured":"Li J, Rakin AS, He Z, Fan D, Chakrabarti C (2021) Radar: run-time adversarial weight attack detection and accuracy recovery. In: 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE). https:\/\/doi.org\/10.23919\/DATE51398.2021.9474113","DOI":"10.23919\/DATE51398.2021.9474113"},{"key":"6693_CR27","doi-asserted-by":"publisher","unstructured":"Javaheripi M, Koushanfar F. Hashtag: hash signatures for online detection of fault-injection attacks on deep neural networks. In: 2021 IEEE\/ACM International Conference on Computer Aided Design (ICCAD). https:\/\/doi.org\/10.1109\/ICCAD51958.2021.9643556","DOI":"10.1109\/ICCAD51958.2021.9643556"},{"key":"6693_CR28","unstructured":"Onnx: Model Zoo Github repository (2023). https:\/\/github.com\/onnx\/models"},{"key":"6693_CR29","unstructured":"Torchvision: pre-treined models repository (2023). https:\/\/pytorch.org\/vision\/stable\/models.html"},{"key":"6693_CR30","unstructured":"Pytorch: export a PyTorch model to ONNX (2023). https:\/\/pytorch.org\/tutorials\/beginner\/onnx\/export_simple_model_to_onnx_tutorial.html#export-a-pytorch-model-to-onnx"},{"key":"6693_CR31","unstructured":"ImageNet: data set (2023). https:\/\/www.image-net.org\/about.php"},{"issue":"2","key":"6693_CR32","doi-asserted-by":"publisher","first-page":"147","DOI":"10.1002\/j.1538-7305.1950.tb00463.x","volume":"29","author":"RW Hamming","year":"1950","unstructured":"Hamming RW (1950) Error detecting and error correcting codes. Bell Syst Tech J 29(2):147\u2013160. https:\/\/doi.org\/10.1002\/j.1538-7305.1950.tb00463.x","journal-title":"Bell Syst Tech J"},{"key":"6693_CR33","unstructured":"Github: Onnx bit analysis (2023). https:\/\/github.com\/IzanCatalan\/OnnxBitAnalysis"},{"key":"6693_CR34","doi-asserted-by":"publisher","unstructured":"Leveugle R, Calvez A, Maistri P, Vanhauwaert P (2009) Statistical fault injection: quantified error and confidence. In: 2009 Design, Automation Test in Europe Conference Exhibition, pp 502\u2013506. https:\/\/doi.org\/10.1109\/DATE.2009.5090716","DOI":"10.1109\/DATE.2009.5090716"},{"key":"6693_CR35","doi-asserted-by":"publisher","unstructured":"Qutub S, Geissler F, Peng Y, Gr\u00e4fe R, Paulitsch M, Hinz G, Knoll A (2022) Hardware faults that matter: understanding and estimating the safety impact of hardware faults on object detection DNNs. In: Computer Safety, Reliability, and Security: 41st International Conference, SAFECOMP 2022, Munich, Germany, September 6\u20139, Proceedings, pp 298\u2013318. Springer, Berlin (2022). https:\/\/doi.org\/10.1007\/978-3-031-14835-4_20","DOI":"10.1007\/978-3-031-14835-4_20"},{"key":"6693_CR36","doi-asserted-by":"publisher","DOI":"10.1109\/DAC.2018.8465834","volume-title":"Ares: a framework for quantifying the resilience of deep neural networks","author":"B Reagen","year":"2018","unstructured":"Reagen B, Gupta U, Pentecost L, Whatmough P, Lee SK, Mulholland N, Brooks D, Wei G-Y (2018) Ares: a framework for quantifying the resilience of deep neural networks. Association for Computing Machinery, New York. https:\/\/doi.org\/10.1109\/DAC.2018.8465834"},{"key":"6693_CR37","unstructured":"Onnx: runtime cross-platform inference and training machine-learning accelerator (2023). https:\/\/onnxruntime.ai\/"},{"key":"6693_CR38","unstructured":"Apache: MxNet library for deep learning (2023). https:\/\/mxnet.apache.org\/versions\/1.9.1\/"},{"key":"6693_CR39","unstructured":"Onnx: API reference (2023). https:\/\/onnx.ai\/onnx\/api\/"},{"key":"6693_CR40","unstructured":"Intel: open vino model zoo repository (2023). https:\/\/docs.openvino.ai\/2023.2\/model_zoo.html"},{"key":"6693_CR41","doi-asserted-by":"publisher","unstructured":"Jouppi NP, Hyun Yoon D, Ashcraft M, Gottscho M, Jablin TB, Kurian G, Laudon J, Li S, Ma P, Ma X, Norrie T, Patil N, Prasad S, Young C, Zhou Z, Patterson D (2021) Ten lessons from three generations shaped Google\u2019s tpuv4i : industrial product. In: 2021 ACM\/IEEE 48th Annual International Symposium on Computer Architecture (ISCA), pp 1\u201314. https:\/\/doi.org\/10.1109\/ISCA52012.2021.00010","DOI":"10.1109\/ISCA52012.2021.00010"},{"key":"6693_CR42","unstructured":"Kalamkar DD, Mudigere D, Mellempudi N, Das D, Banerjee K, Avancha S, Vooturi DT, Jammalamadaka N, Huang J, Yuen H, Yang J, Park J, Heinecke A, Georganas E, Srinivasan S, Kundu A, Smelyanskiy M, Kaul B, Dubey P (2019) A study of BFLOAT16 for deep learning training. CoRR abs\/1905.12322"},{"key":"6693_CR43","unstructured":"Pytorch: empowering PyTorch on Intel\u00ae Xeon\u00ae Scalable processors with Bfloat16 (2023). https:\/\/pytorch.org\/blog\/empowering-pytorch-on-intel-xeon-scalable-processors-with-bfloat16\/"},{"key":"6693_CR44","doi-asserted-by":"publisher","unstructured":"Flich J, Medina L, Catal\u00e1n I, Hern\u00e1ndez C, Bragagnolo A, Auzanneau F, Briand D (2022) Efficient inference of image-based neural network models in reconfigurable systems with pruning and quantization. In: 2022 IEEE International Conference on Image Processing (ICIP), pp 2491\u20132495. https:\/\/doi.org\/10.1109\/ICIP46576.2022.9897752","DOI":"10.1109\/ICIP46576.2022.9897752"},{"key":"6693_CR45","unstructured":"Alveo: U200 data center accelerator card (2023). https:\/\/www.xilinx.com\/products\/boards-and-kits\/alveo"}],"container-title":["The Journal of Supercomputing"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11227-024-06693-7.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s11227-024-06693-7\/fulltext.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11227-024-06693-7.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,11,15]],"date-time":"2024-11-15T12:09:45Z","timestamp":1731672585000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s11227-024-06693-7"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,11,15]]},"references-count":45,"journal-issue":{"issue":"1","published-print":{"date-parts":[[2025,1]]}},"alternative-id":["6693"],"URL":"https:\/\/doi.org\/10.1007\/s11227-024-06693-7","relation":{},"ISSN":["0920-8542","1573-0484"],"issn-type":[{"value":"0920-8542","type":"print"},{"value":"1573-0484","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,11,15]]},"assertion":[{"value":"4 November 2024","order":1,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"15 November 2024","order":2,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"The authors declare no conflict of interest.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Conflict of interest"}}],"article-number":"183"}}