Unveiling the Backdoor Mechanism Hidden Behind Catastrophic Overfitting in Fast Adversarial Training
本文旨在解释快速对抗训练（Fast Adversarial Training, FAT）中出现的灾难性过拟合（Catastrophic Overfitting, CO）现象。FAT能高效提升神经网络对对抗样本的鲁棒性，但容易发生CO，即模型
2026-04-28 归档
今日共收录 1050 条安全情报，包含 258 个 CVE，457 篇安全通告，以及 29 篇研究论文。
← 返回归档列表Mitigating Error Amplification in Fast Adversarial Training
本文针对快速对抗训练（FAT）中存在的两个核心问题展开研究：（1）灾难性过拟合（CO），即模型过度拟合训练时使用的对抗样本，导致对未见攻击泛化能力差；（2）鲁棒性与准确率之间的权衡，即在提高鲁棒性的同时往往导致干净样本上的性能显著下降，且随
RowHammer Vulnerability Counter (RVC): Redefining RowHammer Detection with Victim-Centric Tracking
Rowhammer漏洞随着DRAM工艺进步和代际更迭变得日益严峻，现有缓解方案如Graphene、Twice、Hydra主要依赖跟踪每行的激活次数，并在达到预设阈值时发出刷新指令。然而，这些方法存在固有缺陷，例如难以准确识别真正面临比特翻转
Resolving Conflicts Between RTOS Timekeeping and Uninterruptable Trusted Computing
这篇论文研究了在低功耗微控制器（例如ARM TrustZone-M）上，实时操作系统（RTOS）的时间保持与可信执行环境（TEE）之间的冲突问题。在TrustZone-M架构中，安全世界和非安全世界共享中断控制器等资源。通常，非安全子系统运
Converging Zero Trust and IoT Security: A Multivocal Literature Review
该论文针对物联网（IoT）安全与零信任（ZT）原则的融合趋势，首次采用多声部文献综述（MLR）方法，系统整合了68篇学术文献和36篇行业报告。研究发现，学术界与工业界在零信任与IoT安全融合的路径上存在互补但分歧的视角：学术界主要关注通过修
Detecting Avalanche Effect in Adversarial Settings: Spotting the Encryption Loops in Ransomware
Spotting encryption loops in binary-only ransomware is a critical reverse engineering task. Since the existence of avala
Evaluating Cryptographic API Misuse Detectors for Go
Cryptographic API misuse represents a critical vulnerability class that undermines the security foundations of modern so
The Kerimov-Alekberli Model: An Information-Geometric Framework for Real-Time System Stability
This study introduces the Kerimov-Alekberli model, a novel information-geometric framework that redefines AI safety by f
Jailbreaking Frontier Foundation Models Through Intention Deception
Large (vision-)language models exhibit remarkable capability but remain highly susceptible to jailbreaking. Existing saf
Listen to the Voices of Everyday Users: Democratizing Privacy Ratings for Sensitive Data Access in Mobile Apps
Mobile apps frequently request excessive data access, raising significant privacy concerns. While regulations like GDPR