cs.CR · 2026-11-02 · No. 161
Cryptography and Security, 2026-11-02.
7 new papers in cs.CR. Titles, authors,
abstracts. Links to arXiv. Want this in your inbox every morning? Subscribe →
01 — The papers
7 entries-
01
SecPO: Principled Adversarial Training for Prompt Injection Security
Sizhe Chen, Yibo Peng, Jaewon Chang, Chawin Sitawarin, David Wagner
cs.CR · cs.AI
Recognized as a top threat to AI agents, prompt injection allows instructions from untrusted data (email, documents, websites) to manipulate the agent's operation. The existing best-defended LLM, Meta-SecAlign, is trained with the DPO loss but still suffers from attack success rates (ASRs) approaching 100% under adaptive prompt injections. One reason is that DPO anchors the model around the undefended reference model, which is insecure. We...
-
02
What Standard Evaluations Hide About Detectors of Hijacked LLM Agents: Rare Attacks, Unseen Tools and Benchmark Transfer
Mahmoud Abbasi, Juan M. Núñez V., Amin Shahraki
cs.CR · cs.LG
Detectors can flag hijacked large language model (LLM) agents from the traces of the tools they call. In the eleven evaluations we audited, attacks or anomalies make up 16% to 86% of the test samples, no evaluation varies that share, and six report no test on data from a second benchmark or source. We ask what such evaluations hide. On the AgentDrift and AgentDojo benchmarks we test trained detectors that read tool identities, behavior, call...
-
03
From Training Exposure to Membership Evidence: A Controlled Study of OLMo 2
Ali Banaei, Muhammad Shahab Shahzad, Gerald Q. Maguire, Marco Chiesa, Dejan Kostić
cs.CR · cs.CL · cs.LG
Whether a language model was trained on a document matters for copyright, privacy, and benchmark contamination. Yet most membership-inference (MI) evaluations on language models collect members and non-members after the fact, so success may reflect distribution shift and failure is inconclusive. We re-ran OLMo 2 1B's official 50-billion-token mid-training stage, replacing selected training chunks with documents from 7,814 pairs matched on...
-
04
Simple Features Beat Most AI Control Monitors
Shubhrangshu Debsarkar, Valmik Nahata
cs.CR · cs.AI
AI control relies on monitors to catch a capable model that sabotages its work. These monitors are evaluated on their ability to differentiate honest from backdoored code. We show that on a standard evaluation dataset, simple features beat most monitors without ever running the code, outranking 20 of 24 open-weight models prompted as monitors. The dataset gives its labels away in two ways. People wrote the honest solutions and older language...
-
05
DeepWeave: Weaving Neural and Symbolic Evidence for ML-Based Network Intrusion Detection System
Raoul Saad, Enkeleda Bardhi, Andrea Agiollo, Fernando Kuipers
cs.CR · cs.LG · cs.NI
Machine learning-based network intrusion detection systems can recognize complex traffic patterns, but often require substantial labelled attack data, degrade under deployment shift, and provide limited evidence for analysts. We present DeepWeave, a probabilistic neuro-symbolic Network Intrusion Detection System (NIDS) that makes logic an integral part of detection. DeepWeave automatically constructs and validates interpretable rules,...
-
06
When Safety Checks Become Privacy Risks: A Privacy-Preserving Paradigm for LLM Agent Safety Auditing
Yixiang Luo, Ziheng Ma, Ke Chang, Jian Wang, Jiqiang Liu
cs.CR · cs.AI
Large language model agents interact with online services through tool calls that may access private data or perform sensitive actions. Specialized providers can supply and maintain pre-execution safety auditing capabilities for agent operators. However, sending task descriptions, call arguments, and interaction histories to an external auditor may expose sensitive user information. In this work, we propose a paradigm for privacy-preserving...
-
07
Erase the Injected Region: Multimodal Prompt-Injection Defense with On-Policy Self-Distillation
Shuo Hong, Jiaxin Ma, Xidi Cai, Qianli Ma
cs.CR · cs.AI
Multimodal indirect prompt injection embeds malicious instructions in external media that also contains information needed for the user's task. Existing defenses can block the attack yet impair the model's use of that information, leaving a gap between attack defense and task preservation. We propose ERASE, an on-policy self-distillation method that constructs its reference by erasing the injected region in place. Modality-appropriate blank...
This edition is part of The Daily Abstract — cs.CR archive. Subscribe to receive these in your inbox each morning, automatically translated to Spanish, with reply-to-PDF: arxivdaily.ignorelist.com.
#D99C5E. Built and served on an always-free VM. The masthead is set 14% letterspaced because newspapers do that and it works.