THE DAILY EDITION

AI

6 items · 13 articles in this topic

New models and policy pressure share the day with a steady flow of research on AI agents attacking software. Details on Qwen4-Max are thin in the material.

  1. 1

    Alibaba's Qwen team published a blog post announcing Qwen4-Max, a new flagship model. The summary gives no benchmarks or availability details.

  2. 2

    Anthropic's Claude Fable 5.2 blocks fewer security tasks, while the partner-only Mythos 5.2 gets better at finding bugs, per BleepingComputer.

    NotableBleepingComputer ↗2 min read
  3. 3

    Senator Maria Cantwell criticized Commerce Secretary Howard Lutnick for failing to enforce rules that keep Nvidia AI chips from reaching China.

    NotableThe Verge ↗3 min read
  4. 4

    Apple launched the Apple Health AI Usage Study with OpenEvidence, using an iOS app where users rate AI medical answers.

    NotableThe Verge ↗2 min read
  5. 5

    Several papers target offensive-security agents: HackWorld benchmarks computer-use agents exploiting web app vulnerabilities, CTF-Dojo supplies 658 containerized CTF challenges for training, and Cyber-Zero trains agents from public CTF writeups without a runtime.

  6. 6

    A Lobsters post argues that all AI-generated code should be released as open source.

    NotableLobsters ↗1 min read
More in AI · 7 articles›
  • Cyber-Zero trains cybersecurity LLM agents from public CTF writeups without requiring executable runtime environments.

    arXiv cs.CR (Cryptography & Security)llmcybersecurityagents
    Feed excerpt

    arXiv:2508.00910v3 Announce Type: replace Abstract: Large Language Models (LLMs) have achieved remarkable success in software engineering tasks when trained with executable runtime environments, particularly in resolving GitHub issues. However, such runtime environments are often unavailable in other domains, especially cybersecurity, where challenge configurations and execution contexts are ephemeral or restricted. We present Cyber-Zero, the first runtime-free framework for synthesizing high-quality agent trajectories to train cybersecurity LLMs. Cyber-Zero leverages publicly available CTF writeups and employs persona-driven LLM simulation to reverse-engineer runtime behaviors and generate realistic, long-horizon interaction sequences without actual environments. Using trajectories synthesized by Cyber-Zero, we train LLM-based agents that achieve up to 11.4% absolute performance gains over baseline models on three prominent CTF benchmarks: InterCode-CTF, NYU CTF Bench, and Cybench. Our best model, Cyber-Zero-32B, establishes new state-of-the-art performance among open-weight models, matching the capabilities of proprietary systems like DeepSeek-V3-0324 and Claude-3.5-Sonnet while …

  • Argues that LLM scaling laws fail for security tasks because anomalous long-tail inputs are poorly served by larger models.

    arXiv cs.CR (Cryptography & Security)llmsecurityscaling
    Feed excerpt

    arXiv:2611.17918v1 Announce Type: new Abstract: Scaling laws indicate that the generalization error of a large language model (LLM) decreases monotonically with model size once data is sufficient, and the corollary that larger models outperform smaller ones is now widely accepted. We argue that this corollary fails in security: decisions there hinge on anomalous inputs, which are long-tailed (e.g., zero-day attacks) and therefore poorly served by likelihood fitting. Worse still, the capability brought about by scale (e.g., inferring the intended whole from a corrupted part) leads a LLM silently \emph{repair} the anomaly before any downstream security head ever sees it. We formalize this as a tradeoff between two quantities: the cognitive capability $K(W)=I(W;P)$ that a model holds about the world distribution, and the security-information retention $R(W)=\frac{I(Z;S)}{I(A;S)}$ that survives in its intermediate representation. We derive three information-theoretic bounds to connect them: an upper bound on restoration accuracy in terms of $K(W)$, an upper bound on $R(W)$ in terms of the restoration error, and a Fano lower bound on security-task error in terms of $R(W)$. These theorem…

  • Presents LinEra, a forward-secure erasure scheme ensuring deleted messages and their derived LLM-agent memories are cryptographically unrecoverable.

    arXiv cs.CR (Cryptography & Security)llm-agentsprivacycryptography
    Feed excerpt

    arXiv:2611.18133v1 Announce Type: new Abstract: Large language model (LLM) agents keep a long-term memory that the model itself writes: summaries, extracted facts and embeddings derived from what users said. Deleting a message is therefore not enough: its derivatives, storage residue and backups survive. We define lineage-complete forward-secure erasure: after a forget request, an adversary that later obtains all storage and the agent's current keys learns nothing about the forgotten input or any memory derived from it. We present LinEra, which encrypts each memory under a key derived from its parents' keys along the derivation graph; source keys come from a puncturable pseudorandom function, so forgetting is a single puncture and every descendant becomes unreadable without being located. A sound lineage rule records as parents everything the writer saw, which confines all influence of a source to its descendants. We prove security and lineage integrity and integrate LinEra with mem0. With real LLMs from two families (1.5B to 7B parameters) writing the memory of LoCoMo conversations, an attacker holding every file and the current keys recovers, on all ten conversations written by t…

  • Study compares latent representational geometry of MNIST MLPs under different adversarial training regimes as a robustness diagnostic.

    arXiv cs.CR (Cryptography & Security)adversarial-robustnessneural-networks
    Feed excerpt

    arXiv:2611.17684v1 Announce Type: cross Abstract: Adversarial training is a common technique for making neural networks more robust to small, intentionally crafted input perturbations, and robustness is commonly measured through accuracy under attack. By exploring inter-network representations across different adversarial training objectives, we can analyze whether multiple networks organize their latent representations similarly. We study Multilayer Perceptrons trained on MNIST under three adversarial training regimes and compare the resulting population training dynamics and geometry to those of a vanilla network. We use M-PHATE to embed data samples across training epochs and examine how the geometry of the data changes during training under adversarial objectives. On penultimate-layer representations we measure displacement between clean and perturbed points, distributional distance through maximum mean discrepancy, and nearest-neighbor recovery. Our visualizations and quantitative analyses show that adversarial training leads to structural changes in how networks organize internal data representations, and robust accuracy and representational stability can rank the same defens…

  • Introduces lightweight adapters that transform neural image watermarking models between watermarking styles without full retraining.

    arXiv cs.CR (Cryptography & Security)watermarkingneural-networks
    Feed excerpt

    arXiv:2611.18363v1 Announce Type: cross Abstract: We explore the adaptability of neural image watermarking models; the extent to which one watermarking method can be transformed into another through lightweight adaptation. We introduce a framework that enables encoder-decoder watermarking architectures to become modular and upgradable, supporting multiple watermarking styles without re-training full models. Adaptation is performed by attaching a compact parameter-efficient adapter and fine-tuning under a multi-domain residual alignment loss (spatial, gradient, and frequency) while preserving bit accuracy via a secret-loss term. We evaluate this approach across a suite of recent watermarking models, including TrustMark, InvisMark, RoSteALS, HiDDeN, and WAM, demonstrating that adapted encoders closely reproduce target residuals and that adapted decoders recover messages with near-optimal accuracy under standard augmentations and distortions. We show that cross-model adaptation, with applications in interoperable and upgradable watermarking, can be achieved with as few as 1-2M trainable parameters per component (e.g., the 2M decoder adapter used in our multi-decoder inference pipeline…

  • CTF-Dojo introduces 658 containerized capture-the-flag challenges as an executable runtime for training LLM agents with verifiable feedback.

    arXiv cs.CR (Cryptography & Security)llmagentsctfarxiv
    Feed excerpt

    arXiv:2508.18370v3 Announce Type: replace-cross Abstract: Large language models (LLMs) have demonstrated exceptional capabilities when trained within executable runtime environments, notably excelling at software engineering tasks through verified feedback loops. Yet, scalable and generalizable execution-grounded environments remain scarce, limiting progress in training more capable ML agents. We introduce CTF-Dojo, the first large-scale executable runtime tailored for training LLMs with verifiable feedback, featuring 658 fully functional Capture-The-Flag (CTF)-style challenges containerized in Docker with guaranteed reproducibility. To enable rapid scaling without manual intervention, we develop CTF-Forge, an automated pipeline that transforms publicly available artifacts into ready-to-use execution environments in minutes, eliminating weeks of expert configuration traditionally required. We trained LLM-based agents on just 486 high-quality, execution-verified trajectories from CTF-Dojo, achieving up to 9.9% absolute gains over strong baselines across three competitive benchmarks: InterCode-CTF, NYU CTF Bench, and Cybench. Our best-performing 32B model reaches 31.9% Pass@1, establ…

  • REFORGE is a benchmark method for measuring LLMs' ability to name functions in decompiled binaries, accounting for reliably evaluable ground truth.

    arXiv cs.CR (Cryptography & Security)llmreverse-engineeringbenchmarkarxiv
    Feed excerpt

    arXiv:2607.07738v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly applied to reverse-engineering tasks, and recent threat-intelligence reporting shows them operating inside live offensive-security workflows. Claims about their capability, however, outpace our ability to measure it. Existing benchmarks for LLM-assisted binary analysis treat the construction of function-level ground truth as a solved pre-processing step and report accuracy without disclosing how many functions were reliably evaluable. We argue that the principal obstacle to fair evaluation is not model capability but the reliability of binary-to-source alignment under compiler optimization. This paper presents Reforge, a provenance-tracked pipeline that constructs function-level ground truth from C source through compilation, DWARF and syntactic extraction, alignment, and decompilation, and that operationalizes alignment uncertainty as an eight-gate confidence funnel with three-tier stratification. On a controlled micro-benchmark, high-confidence yield falls from 87.2% to 65.9% across optimization levels, and unpaired comparisons overstate optimization-induced performance decay t…

Written by claude-sonnet-5-5 from feed titles and summaries. The model can misstate details; every claim links to its source.