Mingjie Li, Yukun Jiang, Yicong Tan, Yang Zhang · 2026 · arXiv (Cornell University)
DOI: 10.48550/arxiv.2610.31152 · cited by 0 works (OpenAlex) · open-access PDF
Da AI-Potenzial-Evaluator-Potenzial-Spector-Generierte Bilder in diesen Tagen populär werden, gibt es Diskussionen über solche Generationen in aktuellen öffentlichen Foren, Software-Repositories und Fotografie-Tutorials. Darüber hinaus können bösartige Akteure mit Hilfe von LLM-unterstützten Code-Übersetzungen solche Informationen anwenden, um Detektoren ohne spezielles Wissen oder lokale GPU-Ausführung der resultierenden Tools zu entkommen. Um solche Risiken zu bewerten, führen wir \textsc{WildFakeToolBench} ein, eine Benchmark-Suite von $156 ausführbaren Bildverarbeitungswerkzeugen, die von $220 öffentlichen Quellen über $142 Domains abgeleitet werden. Jedes Tool behält sein Quellbeweis
Machine translation (Meta M2M100); the English original follows.
As AI generated images are getting popular these days, discussions on such generations are widely existed in current public forums, software repositories, and photography tutorials. Moreover, with the help of LLM-assisted code translation, malicious actors can adopt such information to evade detectors without specialized knowledge or local GPU execution of the resulting tools. To evaluate such risks, we introduce \textsc{WildFakeToolBench}, a benchmark suite of $156$ executable image-processing tools derived from $220$ public sources across $142$ domains. Each tool retains its source evidence and passes syntax, execution, and output-distortion checks without using feedback from the evaluated detectors. On $105$ generated images from five generators, we evaluate six state-of-the-art detectors with images processed by the tools in our collected bench to assess their effectiveness against such in-the-wild risks. The strongest individual tools achieve attack success rates of $25.25\%$--$100\%$ across detectors, exceeding the strongest of six selected image-processing baselines for each detector. Evaluating subsets of an eight-tool pool yields worst-case coverage above $80\%$ for every detector. Effectiveness is sparse and detector-specific: high coverage across candidates does not imply reliable one-shot evasion. These findings not only highlight the potential of such web-sourced risks, but also show the potential on establishing AI-image detector robustness evaluation framework based on those in-the-wild information.
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
Topics: Benchmarking, Suite, Computer science, Executable, Detector, Robustness (evolution)
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31152, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange