Hamza Harkous

Wenjun Zeng, Yuchi Liu, Ryan Mullins et al.

We present ShieldGemma, a comprehensive suite of LLM-based safety content moderation models built upon Gemma2. These models provide robust, state-of-the-art predictions of safety risks across key harm types (sexually explicit, dangerous content, harassment, hate speech) in both user input and LLM-generated output. By evaluating on both public and internal benchmarks, we demonstrate superior performance compared to existing models, such as Llama Guard (+10.8\% AU-PRC on public benchmarks) and WildCard (+4.3\%). Additionally, we present a novel LLM-based data curation pipeline, adaptable to a variety of safety-related tasks and beyond. We have shown strong generalization performance for model trained mainly on synthetic data. By releasing ShieldGemma, we provide a valuable resource to the research community, advancing LLM safety and enabling the creation of more effective content moderation solutions for developers.

Kassem Fawaz, Ren Yi, Octavian Suciu et al.

The ability to simulate human privacy decisions has significant implications for aligning autonomous agents with individual intent and conducting cost-effective, large-scale privacy-centric user studies. Prior approaches prompt Large Language Models (LLMs) with natural language user statements, data-sharing histories, or demographic attributes to simulate privacy decisions. These approaches, however, fail to balance individual-level accuracy, prompt usability, token efficiency, and population-level representation. We present Narriva, an approach that generates text-based synthetic privacy personas to address these shortcomings. Narriva grounds persona generation in prior user privacy decisions, such as those from large-scale survey datasets, rather than purely relying on demographic stereotypes. It compresses this data into concise, human-readable summaries structured by established privacy theories. Through benchmarking across five diverse datasets, we analyze the characteristics of Narriva's synthetic personas in modeling both individual and population-level privacy preferences. We find that grounding personas in past privacy behaviors achieves up to 88% predictive accuracy (significantly outperforming a non-personalized LLM baseline), and yields an 80-95% reduction in prompt tokens compared to in-context learning with raw examples. Finally, we demonstrate that personas synthesized from a single survey can reproduce the aggregate privacy behaviors and statistical distributions (TVComplement up to 0.85) of entirely different studies.

Sadegh Momeni, Ge Zhang, Birkett Huber et al.

Despite advancements in machine learning for security, rule-based detection remains prevalent in Security Operations Centers due to the resource intensiveness and skill gap associated with ML solutions. While traditional rule-based methods offer efficiency, their rigidity leads to high false positives or negatives and requires continuous manual maintenance. This paper proposes a novel, two-stage hybrid framework to democratize ML-based threat detection. The first stage employs intentionally loose YARA rules for coarse-grained filtering, optimized for high recall. The second stage utilizes an ML classifier to filter out false positives from the first stage's output. To overcome data scarcity, the system leverages Simula, a seedless synthetic data generation framework, enabling security analysts to create high-quality training datasets without extensive data science expertise or pre-labeled examples. A continuous feedback loop incorporates real-time investigation results to adaptively tune the ML model, preventing rule degradation. This proposed model with active learning has been rigorously tested for a prolonged time in a production environment spanning tens of thousands of systems. The system handles initial raw log volumes often reaching 250 billion events per day, significantly reducing them through filtering and ML inference to a handful of daily tickets for human investigation. Live experiments over an extended timeline demonstrate a general improvement in the model's precision over time due to the active learning feature. This approach offers a self-sustained, low-overhead, and low-maintenance solution, allowing security professionals to guide model learning as expert ``teachers''.

Omer Akgul, Sai Teja Peddinti, Nina Taft et al.

We present an analysis of 12 million instances of privacy-relevant reviews publicly visible on the Google Play Store that span a 10 year period. By leveraging state of the art NLP techniques, we examine what users have been writing about privacy along multiple dimensions: time, countries, app types, diverse privacy topics, and even across a spectrum of emotions. We find consistent growth of privacy-relevant reviews, and explore topics that are trending (such as Data Deletion and Data Theft), as well as those on the decline (such as privacy-relevant reviews on sensitive permissions). We find that although privacy reviews come from more than 200 countries, 33 countries provide 90% of privacy reviews. We conduct a comparison across countries by examining the distribution of privacy topics a country's users write about, and find that geographic proximity is not a reliable indicator that nearby countries have similar privacy perspectives. We uncover some countries with unique patterns and explore those herein. Surprisingly, we uncover that it is not uncommon for reviews that discuss privacy to be positive (32%); many users express pleasure about privacy features within apps or privacy-focused apps. We also uncover some unexpected behaviors, such as the use of reviews to deliver privacy disclaimers to developers. Finally, we demonstrate the value of analyzing app reviews with our approach as a complement to existing methods for understanding users' perspectives about privacy

Tim R. Davidson, Benoit Seguin, Enrico Bacis et al.

Although many AI applications of interest require specialized multi-modal models, relevant data to train such models is inherently scarce or inaccessible. Filling these gaps with human annotators is prohibitively expensive, error-prone, and time-consuming, leading model builders to increasingly consider synthetic data as a scalable alternative. However, existing synthetic data generation methods often rely on manual prompts, evolutionary algorithms, or extensive seed data from the target distribution - limiting their scalability, explainability, and control. In this paper, we introduce Simula: a novel reasoning-driven framework for data generation and evaluation. It employs a seedless, agentic approach to generate synthetic datasets at scale, allowing users to define desired dataset characteristics through an explainable and controllable process that enables fine-grained resource allocation. We show the efficacy of our approach on a variety of datasets, rigorously testing both intrinsic and downstream properties. Our work (1) offers guidelines for synthetic data mechanism design, (2) provides insights into generating and evaluating synthetic data at scale, and (3) unlocks new opportunities for developing and deploying AI in domains where data scarcity or privacy concerns are paramount.

Hamza Harkous, Isabel Groves, Amir Saffari

End-to-end neural data-to-text (D2T) generation has recently emerged as an alternative to pipeline-based architectures. However, it has faced challenges in generalizing to new domains and generating semantically consistent text. In this work, we present DataTuner, a neural, end-to-end data-to-text generation system that makes minimal assumptions about the data representation and the target domain. We take a two-stage generation-reranking approach, combining a fine-tuned language model with a semantic fidelity classifier. Each of our components is learnt end-to-end without the need for dataset-specific heuristics, entity delexicalization, or post-processing. We show that DataTuner achieves state of the art results on the automated metrics across four major D2T datasets (LDC2017T10, WebNLG, ViGGO, and Cleaned E2E), with a fluency assessed by human annotators nearing or exceeding the human-written reference texts. We further demonstrate that the model-based semantic fidelity scorer in DataTuner is a better assessment tool compared to traditional, heuristic-based measures. Our generated text has a significantly better semantic fidelity than the state of the art across all four datasets

Thomas Linden, Rishabh Khandelwal, Hamza Harkous et al.

The EU General Data Protection Regulation (GDPR) is one of the most demanding and comprehensive privacy regulations of all time. A year after it went into effect, we study its impact on the landscape of privacy policies online. We conduct the first longitudinal, in-depth, and at-scale assessment of privacy policies before and after the GDPR. We gauge the complete consumption cycle of these policies, from the first user impressions until the compliance assessment. We create a diverse corpus of two sets of 6,278 unique English-language privacy policies from inside and outside the EU, covering their pre-GDPR and the post-GDPR versions. The results of our tests and analyses suggest that the GDPR has been a catalyst for a major overhaul of the privacy policies inside and outside the EU. This overhaul of the policies, manifesting in extensive textual changes, especially for the EU-based websites, comes at mixed benefits to the users. While the privacy policies have become considerably longer, our user study with 470 participants on Amazon MTurk indicates a significant improvement in the visual representation of privacy policies from the users' perspective for the EU websites. We further develop a new workflow for the automated assessment of requirements in privacy policies. Using this workflow, we show that privacy policies cover more data practices and are more consistent with seven compliance requirements post the GDPR. We also assess how transparent the organizations are with their privacy practices by performing specificity analysis. In this analysis, we find evidence for positive changes triggered by the GDPR, with the specificity level improving on average. Still, we find the landscape of privacy policies to be in a transitional phase; many policies still do not meet several key GDPR requirements or their improved coverage comes with reduced specificity.

Hamza Harkous, Kassem Fawaz, Rémi Lebret et al.

Privacy policies are the primary channel through which companies inform users about their data collection and sharing practices. These policies are often long and difficult to comprehend. Short notices based on information extracted from privacy policies have been shown to be useful but face a significant scalability hurdle, given the number of policies and their evolution over time. Companies, users, researchers, and regulators still lack usable and scalable tools to cope with the breadth and depth of privacy policies. To address these hurdles, we propose an automated framework for privacy policy analysis (Polisis). It enables scalable, dynamic, and multi-dimensional queries on natural language privacy policies. At the core of Polisis is a privacy-centric language model, built with 130K privacy policies, and a novel hierarchy of neural-network classifiers that accounts for both high-level aspects and fine-grained details of privacy practices. We demonstrate Polisis' modularity and utility with two applications supporting structured and free-form querying. The structured querying application is the automated assignment of privacy icons from privacy policies. With Polisis, we can achieve an accuracy of 88.4% on this task. The second application, PriBot, is the first freeform question-answering system for privacy policies. We show that PriBot can produce a correct answer among its top-3 results for 82% of the test questions. Using an MTurk user study with 700 participants, we show that at least one of PriBot's top-3 answers is relevant to users for 89% of the test questions.

Amit Gupta, Rémi Lebret, Hamza Harkous et al.

We propose a novel, semi-supervised approach towards domain taxonomy induction from an input vocabulary of seed terms. Unlike all previous approaches, which typically extract direct hypernym edges for terms, our approach utilizes a novel probabilistic framework to extract hypernym subsequences. Taxonomy induction from extracted subsequences is cast as an instance of the minimumcost flow problem on a carefully designed directed graph. Through experiments, we demonstrate that our approach outperforms stateof- the-art taxonomy induction approaches across four languages. Importantly, we also show that our approach is robust to the presence of noise in the input vocabulary. To the best of our knowledge, no previous approaches have been empirically proven to manifest noise-robustness in the input vocabulary.

Amit Gupta, Rémi Lebret, Hamza Harkous et al.

We propose a simple, yet effective, approach towards inducing multilingual taxonomies from Wikipedia. Given an English taxonomy, our approach leverages the interlanguage links of Wikipedia followed by character-level classifiers to induce high-precision, high-coverage taxonomies in other languages. Through experiments, we demonstrate that our approach significantly outperforms the state-of-the-art, heuristics-heavy approaches for six languages. As a consequence of our work, we release presumably the largest and the most accurate multilingual taxonomic resource spanning over 280 languages.

Hamza Harkous, Karl Aberer

Cloud storage services, like Dropbox and Google Drive, have growing ecosystems of 3rd party apps that are designed to work with users' cloud files. Such apps often request full access to users' files, including files shared with collaborators. Hence, whenever a user grants access to a new vendor, she is inflicting a privacy loss on herself and on her collaborators too. Based on analyzing a real dataset of 183 Google Drive users and 131 third party apps, we discover that collaborators inflict a privacy loss which is at least 39% higher than what users themselves cause. We take a step toward minimizing this loss by introducing the concept of History-based decisions. Simply put, users are informed at decision time about the vendors which have been previously granted access to their data. Thus, they can reduce their privacy loss by not installing apps from new vendors whenever possible. Next, we realize this concept by introducing a new privacy indicator, which can be integrated within the cloud apps' authorization interface. Via a web experiment with 141 participants recruited from CrowdFlower, we show that our privacy indicator can significantly increase the user's likelihood of choosing the app that minimizes her privacy loss. Finally, we explore the network effect of History-based decisions via a simulation on top of large collaboration networks. We demonstrate that adopting such a decision-making process is capable of reducing the growth of users' privacy loss by 70% in a Google Drive-based network and by 40% in an author collaboration network. This is despite the fact that we neither assume that users cooperate nor that they exhibit altruistic behavior. To our knowledge, our work is the first to provide quantifiable evidence of the privacy risk that collaborators pose in cloud apps. We are also the first to mitigate this problem via a usable privacy approach.

Hamza Harkous, Rameez Rahman, Bojan Karlas et al.

Third party apps that work on top of personal cloud services such as Google Drive and Dropbox, require access to the user's data in order to provide some functionality. Through detailed analysis of a hundred popular Google Drive apps from Google's Chrome store, we discover that the existing permission model is quite often misused: around two thirds of analyzed apps are over-privileged, i.e., they access more data than is needed for them to function. In this work, we analyze three different permission models that aim to discourage users from installing over-privileged apps. In experiments with 210 real users, we discover that the most successful permission model is our novel ensemble method that we call Far-reaching Insights. Far-reaching Insights inform the users about the data-driven insights that apps can make about them (e.g., their topics of interest, collaboration and activity patterns etc.) Thus, they seek to bridge the gap between what third parties can actually know about users and users perception of their privacy leakage. The efficacy of Far-reaching Insights in bridging this gap is demonstrated by our results, as Far-reaching Insights prove to be, on average, twice as effective as the current model in discouraging users from installing over-privileged apps. In an effort for promoting general privacy awareness, we deploy a publicly available privacy oriented app store that uses Far-reaching Insights. Based on the knowledge extracted from data of the store's users (over 115 gigabytes of Google Drive data from 1440 users with 662 installed apps), we also delineate the ecosystem for third-party cloud apps from the standpoint of developers and cloud providers. Finally, we present several general recommendations that can guide other future works in the area of privacy for the cloud.