arXiv subjects

Raluca Ada Popa

Publications and source records attributed to Raluca Ada Popa.

At least 19 recordsLinked to original sources

arXiv2026

Repeat-After-Me: Black-Box Adaptive Visual Prompt Injection

Prompt injection is widely recognized as a major security threat to AI agents that interact with untrusted external data, such as websites, documents, and emails. Prior work has shown that, in the text domain, black-box prompt injection can achieve near-perfect attack success rates (ASRs). In the image domain, however, existing visual prompt injection methods are substantially less effective in attacking frontier commercial VLMs for materially harmful behavior. Achieving such outputs is hard because it requires a long and/or format-compliant target string, such as a precise, parseable native tool call with exact function names and arguments. We present Repeat-After-Me, a black-box adaptive visual prompt injection attack that can reveal personally identifiable information or make malicious tool calls. Across both open-weight and commercial frontier VLMs, including Qwen3.6-27B and GPT-5.5, our method achieves ASRs exceeding 82% and 47%, respectively, under a realistic setting in which the benign user prompt is semantically unrelated to the injected task and does not verbally authorize it. Our optimized injection has non-trivial attack transferability across commercial VLMs and benign samples. We show our attack works in cases where adaptive textual prompt injection fails. In a real-world OpenClaw agent connected to Discord, an untrusted user can use a minimally injected image from our attack to overwrite TOOLS.md, enabling future sensitive behaviors like remote code execution and secret exfiltration. We discuss potential defenses.

cs.CR↗

arXiv2026

Hidden Thoughts Are Not Secret: Reasoning Trace Exposure in LLMs

Reasoning traces have become a valuable form of learning signals for improving and transferring the capabilities of large language models. In particular, detailed traces can help distill reasoning behavior from stronger teacher models into weaker student models. The value of capability transfer has motivated many deployed systems with reasoning models to hide raw internal traces and expose at most summaries and answers to users. As a result, we ask whether such interface-level trace hiding prevents users from obtaining useful reasoning supervision through prompting. We study this question with Reasoning Exposure Prompting (REP), a lightweight in-context elicitation method that uses shadow-model-generated demonstrations wrapped in auxiliary code-like formats to raise user-visible reasoning traces from a victim model. Across the common reasoning dataset, different victim models, and different student model distillation, REP substantially increases similarity between exposed and REP-conditioned internal traces while preserving useful reasoning signals.

cs.AI↗

arXiv2026

Daydreaming: Stealing Hidden Agent Skills through Black-Box Task Interaction

Agent skills bundle instructions, reference data, and executable helpers that let a general agent perform specialized tasks. Hosted providers can keep these files secret while selling access to task results, making the skill itself a valuable target. Existing disclosure defenses can block requests that ask for the skill or reproduce its text, but they cannot block customers from submitting the ordinary tasks the service is built to complete. We present Daydreaming, an execution-only attack that steals a multi-file skill through black-box task interactions. The victim is never asked to reveal the skill or grade a reconstruction. Instead, Daydreaming adaptively creates crafted tasks whose results distinguish possible hidden behaviors. It tests individual behaviors, uses attacker-controlled shadow agents to choose a design, and completes each file using stored victim results and local execution checks. We formalize three nested threat levels of access as Differential, Trace, and Output, and focus on Output, where the attacker sees only the final response and returned files. Across 7 skills and 4 victim models, Daydreaming recovers 86.8% of the original skill's capability at Output, outperforming SigLeak by almost 4x. It produces installable skills using a median of 32 victim calls per skill even with disclosure defenses enabled. These results show that hiding skill files and filtering direct disclosure do not, by themselves, prevent functional reconstruction through normal use.

cs.CR↗

arXiv2026

The Next Challenge for Agentic Cybersecurity: A Realistic, Contamination-Free Reverse Engineering Benchmark

AI agents are rapidly improving in cybersecurity capabilities when the source code is available for analysis, yet much of the software most consequential to cybersecurity, including malware, firmware, and proprietary applications, is available only as binaries. Analyzing such software requires reverse engineering(RE): recovering program semantics before the analysis can be meaningfully performed. However, evaluating agentic RE poses a fundamental challenge: benchmark instances must be unseen as source code in the LLMs' training data to prevent models from taking shortcuts by recognizing them rather than really analyzing them, while also matching the scale and anti-analysis protections of real software. Unfortunately, however, existing benchmarks do not jointly satisfy these requirements. To this end, we introduce SRE-Bench, the first realistic, contamination-free RE benchmark. Built entirely from scratch by RE experts with over 5,000 hours, SRE-Bench comprises 19 private, real-world-scale programs averaging 16.9K lines of code. We further developed 44 in-house anti-analysis primitives, yielding 262 binary instances and 1572 deterministically graded tasks. Our evaluation across five frontier LLMs (GPT-5.6-sol,Claude-Opus-5,GPT-5.5,Grok-4.5, and GLM-5.2) shows that RE remains largely unsolved: the strongest model, GPT-5.6-sol, scores 61.4% per instance, and fully solves only 31.5% of the instances. Our analysis further reveals that agents behave differently from human engineers, where agents are relatively insensitive to compiler optimization and static linking. Controlled ablations also confirm that both contamination control and realistic scale are essential. These results indicate that strong source-code security capabilities do not yet transfer to binary analysis, highlighting RE as an important frontier for agentic cybersecurity and SRE-Bench as a rigorous testbed to measure progress.

cs.CR↗

arXiv2026

Antiproof: Synthesizing Vulnerability Detectors and Proofs of Exploitability

Discovering vulnerabilities before attackers exploit them requires high recall and reliable automatic validation, but existing approaches struggle to achieve both without prohibitive cost. We present Antiproof, an end-to-end vulnerability discovery system that combines neuro-symbolic detector synthesis for high-recall discovery with proof-of-exploitability oracles for automatic validation. Antiproof learns and iteratively refines static detectors from vulnerability datasets, then validates candidates by verifying whether executable proofs demonstrate concrete attacker capabilities. Evaluated on BountyBench and our curated KEVBench dataset, Antiproof detects 64 of 66 vulnerabilities, improving recall by more than 60 percentage points over static-analysis and neuro-symbolic baselines. In a scan of 50 widely deployed systems, Antiproof uncovered several hundred previously unknown vulnerabilities. We are responsibly disclosing all confirmed zero-days and have received 12 CVE assignments to date, including remote code execution vulnerabilities in Ray, SGLang, vLLM, and LiteLLM that could allow attackers to take over LLM training and inference systems.

cs.CR↗

arXiv2026

Prismata: Confining Cross-Site Prompt Injection in Web Agents

Autonomous web agents promise to automate everyday browsing tasks, but inherit one of the web's oldest attack surfaces. Cross-Site Scripting proved that mixing trusted and untrusted content is dangerous, even on benign pages. Agents resurface this risk by interpreting natural language as instructions, allowing third-party and user-generated content to hijack the agent via prompt injection. The core challenge is that deriving a task-specific security policy requires reasoning over page structure that is entangled with the attacker's content. We present Prismata, a defense enforcing contextual least privilege for web agents, constraining both what the agent sees and what it can do. Prismata's dynamic trust derivation produces permission labels for page content, with structural confinement guarantees, inspired by classical integrity models, that bound any labeling errors so that labels can only decrease in privilege and mislabelings are bounded. Prismata's mechanical confinement enforces these labels by redacting content and restricting agent capabilities. Importantly, these mechanisms require no developer annotations, so Prismata supports the long tail of websites. Across recent published web agent attacks, including adaptive variants, Prismata substantially reduces attack success while preserving benign task utility.

cs.CR↗

arXiv2026

ShannonProver: Towards Automating Formal Cryptographic Proofs

Cryptographic proofs are produced at a scale that increasingly exceeds the community's ability to verify them manually. Machine-checked proofs offer a path toward scalable proof verification, but writing proof scripts for expressive proof assistants such as EasyCrypt remains a major bottleneck: even when the high-level proof plan is known, converting it into proof tactics requires substantial reasoning effort. This paper presents ShannonProver, an agentic framework for automating cryptographic proofs. ShannonProver targets the setting in which a cryptographer provides the security model and a decomposition of the target theorem into lemma-level proof obligations, while the system automatically constructs EasyCrypt proof scripts for those obligations. We evaluate ShannonProver on a dataset of formal cryptographic proofs in EasyCrypt. The dataset spans textbook primitives, deployed protocols, and standardization efforts such as NIST proposals, and includes expert case studies drawn from a corpus that has not previously been available online. We show that ShannonProver can automate substantial portions of cryptographic proof engineering for case studies such as ChaChaPoly1305 and MEE-CBC. More broadly, this work suggests a path toward accelerating cryptographic research: as agents automate the proof-engineering burden, cryptographers can iterate more quickly on new constructions, obtain machine-checked assurance earlier, and bring trustworthy protocols from design to deployment faster.

cs.CR↗

arXiv2026

Chai: Agentic Discovery of Cryptographic Misuse Vulnerabilities

AI-assisted vulnerability discovery has proven effective for bug classes like memory safety, where instrumentation confirms memory violations and efficiently filters false positives. Many dangerous vulnerability classes, such as cryptographic misuse, however, lack any comparable instrumentation. In this work, we present Chai, an AI-based system that discovers and validates cryptographic misuse vulnerabilities through naturally occurring signals. To achieve this, Chai rethinks the classical technique of differential testing by leveraging AI to 1) improve precision for detecting real security issues in libraries, and 2) repurpose commonly overlooked discrepancies as leads for tangible vulnerabilities in downstream applications. In doing so, Chai inverts the prevailing paradigm of AI vulnerability discovery: instead of auditing one codebase for many flaws, it catalogs flaws at the library level and propagates them across a cryptographic dependency graph, delivering compounding efficiency gains. We evaluate Chai across X.509, JWT, and SAML libraries. Chai discovered a previously unknown critical vulnerability in an SSL library that powers billions of devices, along with security bugs in one library behind a major web browser and another in major Linux distributions. In total, these techniques surfaced over 100 vulnerabilities.

cs.CR↗

arXiv2026

GradShield: Alignment Preserving Finetuning

Large Language Models (LLMs) pose a significant risk of safety misalignment after finetuning, as models can be compromised by both explicitly and implicitly harmful data. Even some seemingly benign data can inadvertently steer a model towards misaligned behaviors. To address this, we introduce GradShield, a principled filtering method that safeguards LLMs during finetuning by identifying and removing harmful data points before they corrupt the model's alignment. It removes potentially harmful data by computing a Finetuning Implicit Harmfulness Score (FIHS) for each data point and employs an adaptive thresholding algorithm. We apply GradShield to multiple utility fine-tuning tasks across varying levels of harmful data and evaluate the safety and utility performance of the resulting LLMs using various metrics. The results show that GradShield outperforms all baseline methods, consistently maintaining an Attack Success Rate (ASR) below $6\%$ while preserving utility performance.

cs.CL↗

arXiv2026

Web Agents Should Adopt the Plan-Then-Execute Paradigm

ReAct has become the default architecture across LLM agents, and many existing web agents follow this paradigm. We argue that it is the wrong default for web agents. Instead, web agents should default to plan-then-execute: commit to a task-specific program before observing runtime web content, then execute it. The reason is that web content mixes inputs from many parties. An e-commerce product page may combine a seller's listing, customer reviews and sponsored advertisements. Under ReAct, all of this content flows into the model when deciding on the next action, creating a direct path for prompt injections to steer the agent's control flow. Plan-then-execute changes this boundary: untrusted data may influence values or branches inside a predefined execution graph, but it cannot redefine the user task or cause the model to synthesize new actions at runtime. We analyze WebArena, a popular web agent benchmark, and find that all tasks are compatible with plan-then-execute, while 80% can be completed with a purely programmatic plan, without any runtime LLM subroutine. We identify the main barrier to adopting plan-then-execute on the web: For it to work well, tools must map cleanly to semantic actions, with effects known before execution, so agents have enough information to plan. The web does not naturally expose that interface. Browser tools such as click, type, and scroll have page-dependent meanings. Planning at this layer is near-sighted: the agent can only see actions on the current page, and later actions appear only after it acts. Closing this gap requires typed interfaces that turn website interactions from clicks and keystrokes to task-level operations. This is an infrastructure problem, not a modeling problem. Web tasks do not need reactivity by default; they need typed, complete, auditable website APIs.

cs.CR↗

arXiv2026

Onyx: Cost-Efficient Disk-Oblivious ANN Search

Approximate nearest neighbor (ANN) search in AI systems increasingly handles sensitive data on third-party infrastructure. Trusted execution environments (TEEs) offer protection, but cost-efficient deployments must rely on external SSDs, which leaks user queries through disk access patterns to the host. Oblivious RAM (ORAM) can hide these access patterns but at a high cost; when paired with existing disk-based ANN search techniques, it makes poor use of SSD resources, yielding high latency and poor cost-efficiency. The core challenge for efficient oblivious ANN search over SSDs is balancing both bandwidth and access count. The state-of-the-art ORAM-ANN design minimizes access count at the ANN level and bandwidth at the ORAM level, each trading-off the other, leaving the combined system with both resources overutilized. We propose inverting this design, minimizing bandwidth consumption in the ANN layer and access count in the ORAM layer, since each component is better suited for its new role: ANN's inherent approximation allows for more bandwidth efficiency, while ORAM has no fundamental lower bounds on access count (as opposed to bandwidth). To this end, we propose a cost-efficient approach, Onyx, with two new co-designed components: Onyx-ANNS introduces a compact intermediate representation that proactively prunes the majority of bandwidth-intensive accesses without hurting recall, and Onyx-ORAM proposes a locality-aware shallow tree design that reduces access count while remaining compatible with bandwidth-efficient ORAM techniques. Compared to the state-of-the-art oblivious ANN search system, Onyx achieves $1.7-9.9\times$ lower cost and $2.3-12.3\times$ lower latency.

cs.CR↗

arXiv2026

Auditing Black-Box LLM APIs with a Rank-Based Uniformity Test

As API access becomes a primary interface to large language models (LLMs), users often interact with black-box systems that offer little transparency into the deployed model. To reduce costs or maliciously alter model behaviors, API providers may discreetly serve quantized or fine-tuned variants, which can degrade performance and compromise safety. Detecting such substitutions is difficult, as users lack access to model weights and, in most cases, even output logits. To tackle this problem, we propose a rank-based uniformity test that can verify the behavioral equality of a black-box LLM to a locally deployed authentic model. Our method is accurate, query-efficient, and avoids detectable query patterns, making it robust to adversarial providers that reroute or mix responses upon the detection of testing attempts. We evaluate the approach across diverse threat scenarios, including quantization, harmful fine-tuning, jailbreak prompts, and full model substitution, showing that it consistently achieves superior statistical power over prior methods under constrained query budgets.

cs.CR↗

arXiv2026

Opal: Private Memory for Personal AI

Personal AI systems increasingly retain long-term memory of user activity, including documents, emails, messages, meetings, and ambient recordings. Trusted hardware can keep this data private, but struggles to scale with a growing datastore. This pushes the data to external storage, which exposes retrieval access patterns that leak private information to the application provider. Oblivious RAM (ORAM) is a cryptographic primitive that can hide these patterns, but it requires a fixed access budget, precluding the query-dependent traversals that agentic memory systems rely on for accuracy. We present Opal, a private memory system for personal AI. Our key insight is to decouple all data-dependent reasoning from the bulk of personal data, confining it to the trusted enclave. Untrusted disk then sees only fixed, oblivious memory accesses. This enclave-resident component uses a lightweight knowledge graph to capture personal context that semantic search alone misses and handles continuous ingestion by piggybacking reindexing and capacity management on every ORAM access. Evaluated on a comprehensive synthetic personal-data pipeline driven by stochastic communication models, Opal improves retrieval accuracy by 13 percentage points over semantic search and achieves 29x higher throughput with 15x lower infrastructure cost than a secure baseline. Opal is under consideration for deployment to millions of users at a major AI provider.

cs.CR↗

arXiv2025

MiniScope: A Least Privilege Framework for Authorizing Tool Calling Agents

Tool calling agents are an emerging paradigm in LLM deployment, with major platforms such as ChatGPT, Claude, and Gemini adding connectors and autonomous capabilities. However, the inherent unreliability of LLMs introduces fundamental security risks when these agents operate over sensitive user services. Prior approaches either rely on manually written policies that require security expertise, or place LLMs in the confinement loop, which lacks rigorous security guarantees. We present MiniScope, a framework that enables tool calling agents to operate on user accounts while confining potential damage from unreliable LLMs. MiniScope introduces a novel way to automatically and rigorously enforce least privilege principles by reconstructing permission hierarchies that reflect relationships among tool calls and combining them with a mobile-style permission model to balance security and ease of use. To evaluate MiniScope, we create a synthetic dataset derived from ten popular real-world applications, capturing the complexity of realistic agentic tasks beyond existing simplified benchmarks. Our evaluation shows that MiniScope incurs only 1-6% latency overhead compared to vanilla tool calling agents, while significantly outperforming the LLM based baseline in minimizing permissions as well as computational and operational costs.

cs.CR↗

arXiv2025

Semantic-Aware Parsing for Security Logs

Security logs are foundational to threat detection and post-incident investigation, yet analysts often struggle to fully leverage them due to their heterogeneity and unstructured nature. The standard practice of manually writing parsers to normalize the data in security event management systems is time-consuming and costly due to the long tail of log formats. Meanwhile, querying raw logs without explicit parsing using large language models (LLMs) is impractical at scale. In this paper, we introduce Matryoshka, an end-to-end system leveraging LLMs to automatically generate semantically-aware structured log parsers without labeled examples or human intervention. Matryoshka achieves this by directly inferring log syntax, variable naming, and normalization to common security-specific schemas (e.g., OCSF [1]) from unlabeled log line samples, then generating deterministic parsers and mapping rules that can be efficiently applied during data ingest. This approach provides analysts with semantically-rich data representations at scale, facilitating rapid and precise log search without the traditional burden of manual parser construction. We evaluate Matryoshka's capabilities through both established template generation datasets and new datasets curated to establish end-to-end performance on a realistic distribution of log types. Our experiments show that Matryoshka outperforms prior work on syntax parsing while matching human-generated parsers in both side-by-side comparisons and retrieval for security-relevant queries. These results demonstrate that Matryoshka significantly reduces manual effort by automatically extracting and organizing valuable security data, moving us closer to fully automated, AI-driven analytics.

cs.CR↗

arXiv2025

Budget-aware Test-time Scaling via Discriminative Verification

Test-time scaling is a powerful strategy for boosting the performance of large language models on complex reasoning tasks. While state-of-the-art approaches often employ generative verifiers to select the best solution from a pool of candidates, this method incurs prohibitive computational costs, limiting its practicality. In this work, we shift the focus to a more budget-aware paradigm: discriminative verification. We conduct a thorough empirical analysis and demonstrate that while discriminative verifiers may underperform in isolation, combining them with self-consistency in a hybrid approach creates a powerful and efficient test-time scaling mechanism. Notably, under a fixed compute budget, this hybrid approach surpasses state-of-the-art generative verification by a significant margin: achieving up to 15.3\% higher accuracy on AIME2025. Our findings establish that for practical, real-world applications, budget-aware scaling with discriminative verifiers is not only a "free" upgrade over self-consistency, but also a more effective and efficient alternative to costly generative techniques. Code is available at https://github.com/wang-research-lab/verification.

cs.AI↗

arXiv2025

Membrane: A Cryptographic Access Control System for Data Lakes

Organizations use data lakes to store and analyze sensitive data. But hackers may compromise data lake storage to bypass access controls and access sensitive data. To address this, we propose Membrane, a system that (1) cryptographically enforces data-dependent access control views over a data lake, (2) without restricting the analytical queries data scientists can run. We observe that data lakes, unlike DBMSes, disaggregate computation and storage into separate trust domains, making at-rest encryption sufficient to defend against remote attackers targeting data lake storage, even when running analytical queries in plaintext. This leads to a new system design for Membrane that combines encryption at rest with SQL-aware encryption. Using block ciphers, a fast symmetric-key primitive with hardware acceleration in CPUs, we develop a new SQL-aware encryption protocol well-suited to at-rest encryption. Membrane adds overhead only at the start of an interactive session due to decrypting views, delaying the first query result by up to $\approx 20\times$; subsequent queries process decrypted data in plaintext, resulting in low amortized overhead.

cs.CR↗

arXiv2025

Proof of Sampling: A Nash Equilibrium-Based Verification Protocol for Decentralized Systems

This paper introduces the Proof of Sampling (PoSP) protocol, a Nash Equilibrium-based verification mechanism, and its application to decentralized machine learning inference through spML. Our protocol has a pure strategy Nash Equilibrium, compelling rational participants to act honestly. It economically disincentivizes dishonest behavior, making it costly for participants to compromise the network's integrity. In our spML protocol, we apply PoSP to decentralized inference for AI applications via a novel cryptographic protocol. The resulting protocol is much more efficient than zero knowledge proof based approaches. Moreover, we anticipate that the PoSP protocol could be effectively utilized for designing verification mechanisms within Actively Validated Services (AVS) in restaking solutions. We further expect that the PoSP protocol could be applied to a variety of other decentralized applications. Our approach enhances the reliability and efficiency of decentralized systems, paving the way for a new generation of decentralized applications.

cs.GT↗