https://techgazette.com.br/research/#security This week's papers in Security, each with what its authors did and found, in sentences from its own abstract. en 180 Thu, 15 Oct 2026 17:48:00 GMT https://arxiv.org/abs/2610.22301 https://arxiv.org/abs/2610.22301 Thu, 15 Oct 2026 17:48:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We study the resulting vulnerability class, filesystem capability laundering, where attacker-steered arguments cause a tool to access sensitive host files outside its declared contract and deployment scope. Found: Across 21,097 MCP servers from nine registries, 787 of 4,806 triaged servers (16.4%) exhibit filesystem capabilities despite non-filesystem functional classifications. By Zeqian Huang, Mingyue Tang, Aonan Guan, Qiming Sun, Pengzhou He. https://arxiv.org/abs/2610.22244 https://arxiv.org/abs/2610.22244 Thu, 15 Oct 2026 17:19:00 GMT arXiv cs.CR Security arXiv cs.CR — Did: We build LinguaGuardBench from the 1,220 labelled steps of TS-Bench AgentDojo-Traj: we translate only the user instruction into German, French, Turkish, Swahili and Amharic, and keep every tool call, observation and label byte-identical, so any change in a verdict is attributable to language alone. By Hasan Demirkiran, Jens Ernstberger. https://arxiv.org/abs/2610.22148 https://arxiv.org/abs/2610.22148 Thu, 15 Oct 2026 16:17:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: In this paper, we propose MAppPrint, a robust and real-time app fingerprinting framework. Found: Meanwhile, MAppPrint is robust in dynamic networking conditions, while achieving a throughput of 10.5 GB/s and a prediction latency of 0.68 seconds. By Tao Duan, Junzhou Zhao, Jianglong Li, RongHua Li, Youqian Zhang, Nuo Xu. https://arxiv.org/abs/2610.22047 https://arxiv.org/abs/2610.22047 Thu, 15 Oct 2026 15:16:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We develop a new Hoare logic and type theory to automatically enforce correctness, confidentiality, and integrity properties of protocols written in the Prelude/Overture language framework. By Christian Skalka, Steven Baldasty, Yeonhee Yeh, Felix Walberg, Joseph P. Near. https://arxiv.org/abs/2610.21713 https://arxiv.org/abs/2610.21713 Thu, 15 Oct 2026 11:32:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We study whether item embeddings can reveal item identities when an adversary has access to auxiliary item text from external sources. Found: We show that a simple cosine-similarity attack can recover item identities with high mean reciprocal rank across language models and candidate-set settings on two real-world datasets. By Linda Beková, Rodrigo Alves. https://arxiv.org/abs/2610.21565 https://arxiv.org/abs/2610.21565 Thu, 15 Oct 2026 09:22:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We identify a vulnerability in this design, termed the \emph{Solution Determinacy Property (SDP)}: the obfuscation secret is uniquely determined, exposing the structure of the true key and enabling attackers to break the protection. Found: Experiments across diverse models and datasets show that \textsc{BiShield-TEE} confines attack accuracy to the full-shielding black-box baseline, while achieving higher inference efficiency than prior methods on real hardware.Our code is available at https://github.com/laduiw/BiShield-TEE. By Zhiyuan Wang, Yuezhi Zhou, Ziqi Zhang, Zhu Zhang, Zhixing Tan, Yongheng Deng. https://arxiv.org/abs/2610.21526 https://arxiv.org/abs/2610.21526 Thu, 15 Oct 2026 08:56:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: In this paper, we study the robustness of widely deployed gossip protocols used in committee-based blockchains against realistic network adversaries. Found: Our results reveal surprising tradeoffs: Solana's dynamic trees are bandwidth-efficient in benign settings but fragile under adaptive attacks; Polkadot's grid achieves strong robustness at the cost of high bandwidth overhead; GossipSub is both bandwidth-efficient and resilient to adaptive adversaries, yet highly sensitive to parameter… By Wenting Li, Jannik Albrecht, Frederik Armknecht, Ghassan Karame. https://arxiv.org/abs/2610.21464 https://arxiv.org/abs/2610.21464 Thu, 15 Oct 2026 08:06:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We present a topology-aware topic-level poisoning attack against large-scale HNSW-RAG that generalizes to unseen topic queries. Found: Structural analyses show that the Hub-Satellite topology forms in HNSW, indicating that graph structure contributes to improved poisoned-document retrieval. By Yufei Chen, Yao Wang, Haibin Zhang. https://arxiv.org/abs/2610.21458 https://arxiv.org/abs/2610.21458 Thu, 15 Oct 2026 08:03:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We give the first measurement of marks under this attack, with three evolvers, five proposers and two environments. Found: A scheme-aware plagiarist reshuffles the ties at no measurable cost, and a plagiarist who asks the proposer for more general code removes the key in 75-100% of rewrites. By Jieyu Li, Jingya Li, Sheng Li, Xinpeng Zhang, Zhenxing Qian. https://arxiv.org/abs/2610.21457 https://arxiv.org/abs/2610.21457 Thu, 15 Oct 2026 08:00:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We address this problem by measuring per-object ML-KEM against classical key exchange on an S3-compatible object store, with two client implementations, two processor types and up to 500 concurrent workers, using pre-registered experiments and statistical equivalence tests. Found: Controlled experiments on one processor type traced most of that loss to the way the software library was called, and a second, native client showed that the size of the loss depends on the implementation. By Revanth Reddy Lingala, K. Gangadhara Rao, Padigela Srujan, Saarthak Manocha, Vainavi Venkat. https://arxiv.org/abs/2610.21394 https://arxiv.org/abs/2610.21394 Thu, 15 Oct 2026 07:13:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: Matoi separates reusable consequence evidence from current execution authority and durable occurrence history. Found: Combined representation/reuse reductions against full proof are 10.87 and 16.53 percent; five of twelve family intervals for the matched comparison cross one in each language. By Zexun Wang. https://arxiv.org/abs/2610.21373 https://arxiv.org/abs/2610.21373 Thu, 15 Oct 2026 06:54:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: In this paper, we first conduct an empirical evaluation of mainstream Rust SAST tools using security vulnerabilities identified in real-world Rust applications. By Huaien Zhang, Zhengjie Zhao, Yu Pei, Tian Dong, Minxue Pan, Shin Hwei Tan. https://arxiv.org/abs/2610.21309 https://arxiv.org/abs/2610.21309 Thu, 15 Oct 2026 05:57:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We present LLM4SAST, the first universal framework detecting faults in SAST tools using large language models and dynamic feedback. Found: The proposed heuristics eliminate 47.8% of false positives, and LLM4SAST detects 3.4 times more faults while requiring only 39.8% of the variants compared to existing approaches. By Huaien Zhang, Yu Pei, Hao Chen, Shin Hwei Tan. https://arxiv.org/abs/2610.21276 https://arxiv.org/abs/2610.21276 Thu, 15 Oct 2026 05:11:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: Field nodes run image classifiers locally, exposing a first-class asset, the trained model, to theft. Found: Against the same architecture sliced from a public pretrained backbone, the adversary downloads those stages, recovering at least 90% of maximum accuracy at all depths, taking usable CIFAR-100 seizures from 0.0% to 97.3% at 5 s, given the victim's full labelled training set. By Daniel Gilkarov, Ran Dubin. https://arxiv.org/abs/2610.21269 https://arxiv.org/abs/2610.21269 Thu, 15 Oct 2026 05:03:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: In this paper, we study trigger value selection as an independent design factor in clean-label backdoor attacks against feature-based malware classifiers. Found: These findings show that trigger value selection is not merely an implementation detail, but an important design choice that shapes the effectiveness and detectability profile of clean-label malware backdoor attacks. By Tri Budi Sabdo Utomo, Kazushi Okamoto, Koki Karube, Kei Harada, Atsushi Shibata. https://arxiv.org/abs/2610.21245 https://arxiv.org/abs/2610.21245 Thu, 15 Oct 2026 04:41:00 GMT arXiv (systems & security) cs.CR Security arXiv (systems & security) cs.CR — Did: We analyze their underlying assumptions and privacy--utility trade-offs. By Zhiyuan Zhang.