fault injection testing

Fault injection attacks (FIAs) pose a severe threat to hardware security, yet efficient and systematic pre-silicon evaluation methodologies remain lacking. This work proposes InjectV—the first simulation-level fault injection framework tailored for RISC-V architectures—built upon the gem5 full-system simulator. InjectV enables precise and controllable transient fault injection at security-critical points such as control-flow decisions, counters, and comparisons, covering both registers and memory. By offering a white-box, developer-friendly perspective, the framework significantly enhances vulnerability detection efficiency. Evaluation on the FISSC benchmark suite, including the hardened VerifyPIN variant, demonstrates that InjectV effectively identifies exploitable attack points while reducing assessment time by 95.8% compared to conventional approaches.

This work addresses the lack of systematic analysis and evaluation of fault injection attack techniques. We propose a hierarchical, beginner-oriented taxonomy covering major injection dimensions—including timing, power, electromagnetic, and laser—and unify their attack models and implementation constraints. Through reproducible experiments and multidimensional cost-benefit analysis—assessing hardware overhead, success rate, repeatability, and stealth—we conduct the first cross-tool evaluation of 12 state-of-the-art fault injection methods and platforms. Our findings reveal critical bottlenecks in controllability and automation across existing tools. To address these, we introduce two practical innovations: a lightweight trigger mechanism and a cross-platform validation framework. The study delivers an actionable learning pathway for security practitioners and identifies high-fidelity modeling and AI-driven injection as pivotal research directions for the academic community.

This study addresses the threat posed by single-event upsets (SEUs) to physics-informed Real NVP normalizing flow-based anomaly detection models deployed on orbiting satellites, representing the first systematic radiation resilience assessment of such AI models. Method: We propose a hierarchical state/output dual-path fault injection framework, implemented within a custom TensorFlow environment, enabling multi-granularity (zero-value, random-value, and bit-flip) injection into weights, biases, and activation values. Contribution/Results: Our analysis reveals nonlinear performance degradation patterns triggered by critical-bit flips, identifies radiation-sensitive layers and vulnerable bit positions, and quantifies robustness decay across varying injection locations, fault types, and intensities. The findings provide reproducible empirical evidence and methodological support for fault-tolerant design and high-reliability deployment of onboard AI models.

In distributed systems, self-sustaining cascading failures—arising from multi-step dependency chains and specific environmental conditions—are notoriously difficult to expose prior to deployment. To address this, we propose *Causal Stitching*, a novel paradigm that integrates counterfactual causal analysis (FCA), a three-stage test budget allocation strategy, and local compatibility checking into an end-to-end detection framework. Our approach combines controlled fault injection, execution trace comparison, path-constraint solving, and causal reasoning to precisely identify conditional cascading failure paths. Evaluated on five real-world distributed systems, the framework uncovered 15 self-sustaining cascading failure vulnerabilities; five were officially confirmed by maintainers, and two have already been patched. This significantly enhances pre-deployment exposure of complex, condition-dependent failures and improves the interpretability and diagnostic rigor of root-cause analysis.

Distributed systems operating in complex environments are highly susceptible to failures and adversarial behaviors, making their performance difficult to predict directly from formal designs. This work proposes the first framework that systematically enables performance prediction from formal models by integrating a reusable fault-injection library with a modeling methodology based on the Maude language. The authors develop an automated tool, PERF, which combines model composition and statistical analysis techniques to accurately estimate key performance metrics—such as throughput and latency—across diverse failure scenarios. Experimental evaluation demonstrates that PERF’s predictions align closely with measurements from real-world deployments on representative distributed systems, significantly enhancing the practical utility of formal methods in performance assessment.

This study addresses the limitations of traditional fault injection methods, which test individual layers in isolation and fail to capture cross-layer fault propagation, leaving critical vulnerabilities undetected in multi-layer self-healing AI infrastructures. To overcome this, the authors propose ADA-ST, a novel approach that introduces cross-layer modeling via weighted fault propagation graphs and an adaptive injection mechanism guided by an activity-aware scenario selection algorithm to maximize marginal coverage gain. The method further incorporates Fault-Layer Abstraction Mapping (FLAM) to enable knowledge transfer of fault propagation patterns across hardware generations. Experimental results demonstrate that ADA-ST achieves 100% edge coverage on Alpha, Beta, and Gamma platforms within only 9–12 iterations, with FLAM achieving 96%–100% transfer accuracy. Physical validation confirms all four predicted propagation paths, uncovering four new classes of cross-layer vulnerabilities and critical blind spots in self-healing mechanisms.

Existing SRE benchmark tasks are overly simplified and fail to capture the complexity of fault diagnosis and mitigation in real-world production environments. This work proposes the first high-fidelity, scalable evaluation benchmark for SRE agents, built upon a realistic cloud-native system stack that dynamically simulates operational conditions. The benchmark incorporates a fault injector and noise simulator to support diverse failure modes—including metastable and correlated failures—and provides 90 realistic, challenging tasks. Designed with a modular architecture, it enables continuous extension and adaptation. Experimental results demonstrate significant performance disparities among state-of-the-art AI agents across different fault types, with end-to-end success rates varying by up to 40%, thereby validating the benchmark’s effectiveness and inherent difficulty.

This work addresses the vulnerability of quantized neural networks deployed on edge devices to cumulative weight bit flips induced by radiation, a challenge inadequately captured by existing fault evaluation methods that often neglect realistic memory layouts. To overcome this limitation, the authors propose a GDB-based dynamic binary instrumentation framework that injects bit-level faults directly into the target CPU binary, enabling layer-wise sensitivity profiling without model retraining or source code modification. For the first time, this approach facilitates fine-grained, memory-layout-aware analysis of cumulative fault effects in quantized models under real hardware deployment conditions. Validated across diverse network architectures, quantization schemes, and memory configurations, the method accurately identifies critical vulnerable layers, thereby providing empirical foundations for efficient fault-tolerant design.

This work addresses the challenge of comprehensive safety testing for resource-constrained edge-based autonomous driving vision systems, which struggle to perform real-time evaluation under diverse failure conditions. Existing approaches rely on static datasets or manual fault injection, resulting in insufficient coverage. To overcome this limitation, the authors propose a decoupled offline-online fault injection framework: in the offline phase, large language models generate semantically meaningful failure scenarios, and latent diffusion models synthesize high-fidelity sensor degradation data; in the online phase, lightweight, real-time fault-aware inference is enabled via precomputed lookup tables. This approach uniquely integrates large language models and diffusion models to automatically produce structured, high-fidelity perception faults. Evaluation across 460 scenarios shows that while the system achieves an R² of 0.85 on clean data, generated faults such as fog can increase RMSE by 99% and reduce localization accuracy (error < 0.10) to 31.0%, exposing critical gaps in conventional evaluation protocols for edge deployment.

This work addresses the inefficiencies of manual test planning in large-scale AI data center hardware verification, which suffers from low productivity, insufficient coverage, and poor reusability. The paper proposes the first hardware verification framework that integrates generative AI with multi-agent collaboration to automatically construct structured test plans from self-healing verification documents and bills of materials. The framework enables automated test case generation, intelligent coverage gap closure, and end-to-end traceability. Key technical innovations include input normalization, context-aware component classification, and fault mode synthesis, ensuring high portability across platforms. Experimental results on two production platforms demonstrate coverage improvements of 74.2% and 51.4%, respectively, reducing test plan development time from days to hours while achieving high expert acceptance in novel scenarios.