- 2023-10-23Proj CDeepFuzz Paper Reading: POLYCRUISE: A Cross-Language Dynamic Information Flow Analysis
Abstract本文：PolyCruiseMethod:跨编程语言的holisticdynamicinformationflowanalysis(DIFA)usealightlanguage-specificanalysis和language-agnosticonlinedataflowanalysis来计算symbolicdependencies实验：数据集：PolyBench，包含小中大三种等级的benchmarks效
- 2023-10-04Proj CDeepFuzz Paper Reading: NYX: Greybox Hypervisor Fuzzing using Fast Snapshots and Affine Types
Abstract背景：hypervisor(virtualmachinemonitor,VMM)保障了不同虚拟机之间的安全隔离(securityboundaries)用户：攻击场景：在云服务上运行自身的VMinstances,提升权限本文：Nyx目的：coverageguidedhypervisorfuzzermethod:1.fastsnapshotrestorationmechanism2.mu
- 2023-09-08Proj CDeepFuzz Paper Reading: Metamorphic Testing of Deep Learning Compilers
Abstract背景：CompilingDNNmodelsintohigh-efficiencyexecutablesisnoteasy:thecompilationprocedureofteninvolvesconvertinghigh-levelmodelspecificationsintoseveraldifferentintermediaterepresentations(IR),e.g.,graphIRandoperatorIR,an
- 2023-09-06Proj CDeepFuzz Paper Reading: IvySyn: Automated Vulnerability Discovery in Deep Learning Frameworks
Abstract本文：IvySynTask:discovermemoryerrorvulnerabilitiesinDLframeworksBugType:memorysafetyerrors,fatalruntimeerrorsMethod:利用nativeAPIs中静态写明的类型信息在low-levelkernelcode上执行type-awaremutation-basedfuzzingsynthesizeProofof
- 2023-09-05Proj CDeepFuzz Paper Reading: Framework for Evaluating Faithfulness of Local Explanations
Abstract本文:Task:1.studythefaithfulnessofanexplanationsystemtotheunderlyingpredictionmodelonconsistencyandsufficiency2.introducequantitativemeasuresofconsistencyandsufficiency3.provideestimatorsandsamplecomplexityboundsfo
- 2023-09-04Proj CDeepFuzz Paper Reading: DeepTest: automated testing of deep-neural-network-driven autonomous c
Abstract本文:DeepTestTask:asystematictestingtoolforDNN-drivenvehiclesMethod:generatedtestcaseswithreal-worldchangeslikerain,fog,lightingconditions,etc.maxthenumberofactivatedneuronsGithub:https://github.com/ARiSE-Lab/deepTes
- 2023-09-01Proj CDeepFuzz Paper Reading: Automatic differentiation in PyTorch
Abstract本文：描述automaticdifferentiationmoduleofPyTorch包括：LuaTorch,Chainer,HIPSAutogradTask:Providesahigh-performanceenvironmentondifferentdevices(bothCPUsandGPUs)方法：不用symbolicdifferentiation,而是使用differentiationonpurelyimper
- 2023-08-29Proj CDeepFuzz Paper Reading: Deepxplore: Automated whitebox testing of deep learning systems
Abstract背景：现有的深度学习测试在很⼤程度上依赖于⼿动标记的数据，因此通常⽆法暴露罕⻅输⼊的错误⾏为。本文：DeepXploreTask:awhite-boxframeworktotestDLModels方法：neuroncoveragedifferentialtestingwithmultipleDLsystems(models)joint-optimizationpro