This paper presents a configuration-first framework for evaluating cross-backend compatibility in deep learning systems deployed on CPU, GPU, and compiled runtimes. The framework decouples experiments from code using YAML, supports both library and repository models, and employs a three-tier verification protocol covering tensor-level closeness, activation alignment, and task-level metrics. Through 672 checks across multiple models and tolerance settings, we observe that 72.0% of runs pass, with most discrepancies occurring under stricter thresholds. Our results show that detection models and compiled backends are particularly prone to drift, often due to nondeterministic post-processing. We further demonstrate that deterministic adapters and selective fallbacks can substantially improve agreement without significant performance loss. To our knowledge, this is the first unified framework that systematically quantifies and mitigates cross-backend drift in deep learning, providing a reproducible methodology for dependable deployment across heterogeneous runtimes.

Out-of-distribution (OOD) detection discerns OOD data where the predictor cannot make valid predictions as in-distribution (ID) data, thereby increasing the reliability of open-world classification. However, it is typically hard to collect real out-of-distribution (OOD) data for training a predictor capable of discerning ID and OOD patterns. This obstacle gives rise to data generation-based learning methods, synthesizing OOD data via data generators for predictor training without requiring any real OOD data. Related methods typically pre-train a generator on ID data and adopt various selection procedures to find those data likely to be the OOD cases. However, generated data may still coincide with ID semantics, i.e., mistaken OOD generation remains, confusing the predictor between ID and OOD data. To this end, we suggest that generated data (with mistaken OOD generation) can be used to devise an auxiliary OOD detection task to facilitate real OOD detection. Specifically, we can ensure that learning from such an auxiliary task is beneficial if the ID and the OOD parts have disjoint supports, with the help of a well-designed training procedure for the predictor. Accordingly, we propose a powerful data generation-based learning method named Auxiliary Task-based OOD Learning (ATOL) that can relieve the mistaken OOD generation. We conduct extensive experiments under various OOD detection setups, demonstrating the effectiveness of our method against its advanced counterparts.

We present a framework for automated analysis and categorization of .onion websites in the darkweb to facilitate analyst situational awareness of new content that emerges from this dynamic landscape. Over the last two years, our team has developed a large-scale darkweb crawling infrastructure called OnionCrawler that acquires new onion domains on a daily basis, and crawls and indexes millions of pages from these new and previously known .onion sites. It stores this data into a research repository designed to help better understand Tor’s hidden service ecosystem. The analysis component of our framework is called Automated Tool for Onion Labeling (ATOL), which introduces a two-stage thematic labeling strategy: (1) it learns descriptive and discriminative keywords for different categories, and (2) uses these terms to map onion site content to a set of thematic labels. We also present empirical results of ATOL and our ongoing experimentation with it, as we have gained experience applying it to the entirety of our darkweb repository, now over 70 million indexed pages. We find that ATOL can perform site-level thematic label assignment more accurately than keywordbased schemes developed by domain experts — we expand the analyst-provided keywords using an automatic keyword discovery algorithm, and get 12% gain in accuracy by using a machine learning classification model. We also show how ATOL can discover categories on previously unlabeled onions and discuss applications of ATOL in supporting various analyses and investigations of the darkweb.