Search

It is known that LLMs do hallucinate, that is, they return incorrect information as facts. In this paper, we introduce the possibility to study these hallucinations under a structured form: graphs. Hallucinations in this context are incorrect outputs when prompted for well known graphs from the literature (e.g. Karate club, Les Mis\'erables, graph atlas). These hallucinated graphs have the advantage of being much richer than the factual accuracy -- or not -- of a fact; this paper thus argues that such rich hallucinations can be used to characterize the outputs of LLMs. Our first contribution observes the diversity of topological hallucinations from major modern LLMs. Our second contribution is the proposal of a metric for the amplitude of such hallucinations: the Graph Atlas Distance, that is the average graph edit distance from several graphs in the graph atlas set. We compare this metric to the Hallucination Leaderboard, a hallucination rank that leverages 10,000 times more prompts to obtain its ranking.

Auditors need robust methods to assess the compliance of web platforms with the law. However, since they hardly ever have access to the algorithm, implementation, or training data used by a platform, the problem is harder than a simple metric estimation. Within the recent framework of manipulation-proof auditing, we study in this paper the feasibility of robust audits in realistic settings, in which models exhibit large capacities. We first prove a constraining result: if a web platform uses models that may fit any data, no audit strategy -- whether active or not -- can outperform random sampling when estimating properties such as demographic parity. To better understand the conditions under which state-of-the-art auditing techniques may remain competitive, we then relate the manipulability of audits to the capacity of the targeted models, using the Rademacher complexity. We empirically validate these results on popular models of increasing capacities, thus confirming experimentally that large-capacity models, which are commonly used in practice, are particularly hard to audit robustly. These results refine the limits of the auditing problem, and open up enticing questions on the connection between model capacity and the ability of platforms to manipulate audit attempts., Comment: To appear in the IEEE Conference on Secure and Trustworthy Machine Learning, 2024

Existing work in fairness auditing assumes that each audit is performed independently. In this paper, we consider multiple agents working together, each auditing the same platform for different tasks. Agents have two levers: their collaboration strategy, with or without coordination beforehand, and their strategy for sampling appropriate data points. We theoretically compare the interplay of these levers. Our main findings are that (i) collaboration is generally beneficial for accurate audits, (ii) basic sampling methods often prove to be effective, and (iii) counter-intuitively, extensive coordination on queries often deteriorates audits accuracy as the number of agents increases. Experiments on three large datasets confirm our theoretical results. Our findings motivate collaboration during fairness audits of platforms that use ML models for decision-making., Comment: 14 pages, 7 figures, ECAI

Numerous discussions have advocated the presence of a so called rabbit-hole (RH) phenomenon on social media, interested in advanced personalization to their users. This phenomenon is loosely understood as a collapse of mainstream recommendations, in favor of ultra personalized ones that lock users into narrow and specialized feeds. Yet quantitative studies are often ignoring personalization, are of limited scale, and rely on manual tagging to track this collapse. This precludes a precise understanding of the phenomenon based on reproducible observations, and thus the continuous audits of platforms. In this paper, we first tackle the scale issue by proposing a user-sided bot-centric approach that enables large scale data collection, through autoplay walks on recommendations. We then propose a simple theory that explains the appearance of these RHs. While this theory is a simplifying viewpoint on a complex and planet-wide phenomenon, it carries multiple advantages: it can be analytically modeled, and provides a general yet rigorous definition of RHs. We define them as an interplay between i) user interaction with personalization and ii) the attraction strength of certain video categories, which cause users to quickly step apart of mainstream recommendations made to fresh user profiles. We illustrate these concepts by highlighting some RHs found after collecting more than 16 million personalized recommendations on YouTube. A final validation step compares our automatically-identified RHs against manually-identified RHs from a previous research work. Together, those results pave the way for large scale and automated audits of the RH effect in recommendation systems., Comment: To appear in Social Network Analysis and Mining (Springer)

Recent legislation required AI platforms to provide APIs for regulators to assess their compliance with the law. Research has nevertheless shown that platforms can manipulate their API answers through fairwashing. Facing this threat for reliable auditing, this paper studies the benefits of the joint use of platform scraping and of APIs. In this setup, we elaborate on the use of scraping to detect manipulated answers: since fairwashing only manipulates API answers, exploiting scraps may reveal a manipulation. To abstract the wide range of specific API-scrap situations, we introduce a notion of proxy that captures the consistency an auditor might expect between both data sources. If the regulator has a good proxy of the consistency, then she can easily detect manipulation and even bypass the API to conduct her audit. On the other hand, without a good proxy, relying on the API is necessary, and the auditor cannot defend against fairwashing. We then simulate practical scenarios in which the auditor may mostly rely on the API to conveniently conduct the audit task, while maintaining her chances to detect a potential manipulation. To highlight the tension between the audit task and the API fairwashing detection task, we identify Pareto-optimal strategies in a practical audit scenario. We believe this research sets the stage for reliable audits in practical and manipulation-prone setups., Comment: 18 pages, 7 figures

Recent advances in the fingerprinting of deep neural networks detect instances of models, placed in a black-box interaction scheme. Inputs used by the fingerprinting protocols are specifically crafted for each precise model to be checked for. While efficient in such a scenario, this nevertheless results in a lack of guarantee after a mere modification (like retraining, quantization) of a model. This paper tackles the challenges to propose i) fingerprinting schemes that are resilient to significant modifications of the models, by generalizing to the notion of model families and their variants, ii) an extension of the fingerprinting task encompassing scenarios where one wants to fingerprint not only a precise model (previously referred to as a detection task) but also to identify which model family is in the black-box (identification task). We achieve both goals by demonstrating that benign inputs, that are unmodified images, for instance, are sufficient material for both tasks. We leverage an information-theoretic scheme for the identification task. We devise a greedy discrimination algorithm for the detection task. Both approaches are experimentally validated over an unprecedented set of more than 1,000 networks.

Randomized smoothing is a recent and celebrated solution to certify the robustness of any classifier. While it indeed provides a theoretical robustness against adversarial attacks, the dimensionality of current classifiers necessarily imposes Monte Carlo approaches for its application in practice. This paper questions the effectiveness of randomized smoothing as a defense, against state of the art black-box attacks. This is a novel perspective, as previous research works considered the certification as an unquestionable guarantee. We first formally highlight the mismatch between a theoretical certification and the practice of attacks on classifiers. We then perform attacks on randomized smoothing as a defense. Our main observation is that there is a major mismatch in the settings of the RS for obtaining high certified robustness or when defeating black box attacks while preserving the classifier accuracy., Comment: ICASSP 2022

Algorithmic decision making is now widespread, ranging from health care allocation to more common actions such as recommendation or information ranking. The aim to audit these algorithms has grown alongside. In this paper, we focus on external audits that are conducted by interacting with the user side of the target algorithm, hence considered as a black box. Yet, the legal framework in which these audits take place is mostly ambiguous to researchers developing them: on the one hand, the legal value of the audit outcome is uncertain; on the other hand the auditors' rights and obligations are unclear. The contribution of this paper is to articulate two canonical audit forms to law, to shed light on these aspects: 1) the first audit form (we coin the Bobby audit form) checks a predicate against the algorithm, while the second (Sherlock) is more loose and opens up to multiple investigations. We find that: Bobby audits are more amenable to prosecution, yet are delicate as operating on real user data. This can lead to reject by a court (notion of admissibility). Sherlock audits craft data for their operation, most notably to build surrogates of the audited algorithm. It is mostly used for acts for whistleblowing, as even if accepted as a proof, the evidential value will be low in practice. 2) these two forms require the prior respect of a proper right to audit, granted by law or by the platform being audited; otherwise the auditor will be also prone to prosecutions regardless of the audit outcome. This article thus highlights the relation of current audits with law, in order to structure the growing field of algorithm auditing.

Many defenses have emerged with the development of adversarial attacks. Models must be objectively evaluated accordingly. This paper systematically tackles this concern by proposing a new parameter-free benchmark we coin RoBIC. RoBIC fairly evaluates the robustness of image classifiers using a new half-distortion measure. It gauges the robustness of the network against white and black box attacks, independently of its accuracy. RoBIC is faster than the other available benchmarks. We present the significant differences in the robustness of 16 recent models as assessed by RoBIC., Comment: 4 pages, accepted to ICIP 2021

Shadow banning consists for an online social network in limiting the visibility of some of its users, without them being aware of it. Twitter declares that it does not use such a practice, sometimes arguing about the occurrence of "bugs" to justify restrictions on some users. This paper is the first to address the plausibility or not of shadow banning on a major online platform, by adopting both a statistical and a graph topological approach. We first conduct an extensive data collection and analysis campaign, gathering occurrences of visibility limitations on user profiles (we crawl more than 2.5 million of them). In such a black-box observation setup, we highlight the salient user profile features that may explain a banning practice (using machine learning predictors). We then pose two hypotheses for the phenomenon: i) limitations are bugs, as claimed by Twitter, and ii) shadow banning propagates as an epidemic on user-interactions ego-graphs. We show that hypothesis i) is statistically unlikely with regards to the data we collected. We then show some interesting correlation with hypothesis ii), suggesting that the interaction topology is a good indicator of the presence of groups of shadow banned users on the service., Comment: Appearing in INFOCOM 2021

Machine learning classifiers are critically prone to evasion attacks. Adversarial examples are slightly modified inputs that are then misclassified, while remaining perceptively close to their originals. Last couple of years have witnessed a striking decrease in the amount of queries a black box attack submits to the target classifier, in order to forge adversarials. This particularly concerns the black-box score-based setup, where the attacker has access to top predicted probabilites: the amount of queries went from to millions of to less than a thousand. This paper presents SurFree, a geometrical approach that achieves a similar drastic reduction in the amount of queries in the hardest setup: black box decision-based attacks (only the top-1 label is available). We first highlight that the most recent attacks in that setup, HSJA, QEBA and GeoDA all perform costly gradient surrogate estimations. SurFree proposes to bypass these, by instead focusing on careful trials along diverse directions, guided by precise indications of geometrical properties of the classifier decision boundaries. We motivate this geometric approach before performing a head-to-head comparison with previous attacks with the amount of queries as a first class citizen. We exhibit a faster distortion decay under low query amounts (few hundreds to a thousand), while remaining competitive at higher query budgets., Comment: 8 pages

The concept of explainability is envisioned to satisfy society's demands for transparency on machine learning decisions. The concept is simple: like humans, algorithms should explain the rationale behind their decisions so that their fairness can be assessed. While this approach is promising in a local context (e.g. to explain a model during debugging at training time), we argue that this reasoning cannot simply be transposed in a remote context, where a trained model by a service provider is only accessible through its API. This is problematic as it constitutes precisely the target use-case requiring transparency from a societal perspective. Through an analogy with a club bouncer (which may provide untruthful explanations upon customer reject), we show that providing explanations cannot prevent a remote service from lying about the true reasons leading to its decisions. More precisely, we prove the impossibility of remote explainability for single explanations, by constructing an attack on explanations that hides discriminatory features to the querying user. We provide an example implementation of this attack. We then show that the probability that an observer spots the attack, using several explanations for attempting to find incoherences, is low in practical settings. This undermines the very concept of remote explainability in general.

Neural networks are powering the deployment of embedded devices and Internet of Things. Applications range from personal assistants to critical ones such as self-driving cars. It has been shown recently that models obtained from neural nets can be trojaned ; an attacker can then trigger an arbitrary model behavior facing crafted inputs. This has a critical impact on the security and reliability of those deployed devices. We introduce novel algorithms to detect the tampering with deployed models, classifiers in particular. In the remote interaction setup we consider, the proposed strategy is to identify markers of the model input space that are likely to change class if the model is attacked, allowing a user to detect a possible tampering. This setup makes our proposal compatible with a wide range of scenarios, such as embedded models, or models exposed through prediction APIs. We experiment those tampering detection algorithms on the canonical MNIST dataset, over three different types of neural nets, and facing five different attacks (trojaning, quantization, fine-tuning, compression and watermarking). We then validate over five large models (VGG16, VGG19, ResNet, MobileNet, DenseNet) with a state of the art dataset (VGGFace2), and report results demonstrating the possibility of an efficient detection of model tampering., Comment: In the 30th International Symposium on Software Reliability Engineering (ISSRE 2019)

A recent technical breakthrough in the domain of machine learning is the discovery and the multiple applications of Generative Adversarial Networks (GANs). Those generative models are computationally demanding, as a GAN is composed of two deep neural networks, and because it trains on large datasets. A GAN is generally trained on a single server. In this paper, we address the problem of distributing GANs so that they are able to train over datasets that are spread on multiple workers. MD-GAN is exposed as the first solution for this problem: we propose a novel learning procedure for GANs so that they fit this distributed setup. We then compare the performance of MD-GAN to an adapted version of Federated Learning to GANs, using the MNIST and CIFAR10 datasets. MD-GAN exhibits a reduction by a factor of two of the learning complexity on each worker node, while providing better performances than federated learning on both datasets. We finally discuss the practical implications of distributing GANs., Comment: To be published in IPDPS 2019: the 33rd IEEE International Parallel & Distributed Processing Symposium

The training of deep neural network classifiers results in decision boundaries which geometry is still not well understood. This is in direct relation with classification problems such as so called adversarial examples. We introduce zoNNscan, an index that is intended to inform on the boundary uncertainty (in terms of the presence of other classes) around one given input datapoint. It is based on confidence entropy, and is implemented through sampling in the multidimensional ball surrounding that input. We detail the zoNNscan index, give an algorithm for approximating it, and finally illustrate its benefits on four applications, including two important problems for the adoption of deep networks in critical systems: adversarial examples and corner case inputs. We highlight that zoNNscan exhibits significantly higher values than for standard inputs in those two problem classes.

Recommenders personalize the web content by typically using collaborative filtering to relate users (or items) based on explicit feedback, e.g., ratings. The difficulty of collecting this feedback has recently motivated to consider implicit feedback (e.g., item consumption along with the corresponding time). In this paper, we introduce the notion of consumed item pack (CIP) which enables to link users (or items) based on their implicit analogous consumption behavior. Our proposal is generic, and we show that it captures three novel implicit recommenders: a user-based (CIP-U), an item-based (CIP-I), and a word embedding-based (DEEPCIP), as well as a state-of-the-art technique using implicit feedback (FISM). We show that our recommenders handle incremental updates incorporating freshly consumed items. We demonstrate that all three recommenders provide a recommendation quality that is competitive with state-of-the-art ones, including one incorporating both explicit and implicit feedback., Comment: 12 pages

The state of the art performance of deep learning models comes at a high cost for companies and institutions, due to the tedious data collection and the heavy processing requirements. Recently, [35, 22] proposed to watermark convolutional neural networks for image classification, by embedding information into their weights. While this is a clear progress towards model protection, this technique solely allows for extracting the watermark from a network that one accesses locally and entirely. Instead, we aim at allowing the extraction of the watermark from a neural network (or any other machine learning model) that is operated remotely, and available through a service API. To this end, we propose to mark the model's action itself, tweaking slightly its decision frontiers so that a set of specific queries convey the desired information. In the present paper, we formally introduce the problem and propose a novel zero-bit watermarking algorithm that makes use of adversarial model examples. While limiting the loss of performance of the protected model, this algorithm allows subsequent extraction of the watermark using only few queries. We experimented the approach on three neural networks designed for image classification, in the context of MNIST digit recognition task., Comment: To appear in the journal of Neural Computing and Applications, 2019

Recommendation plays a key role in e-commerce and in the entertainment industry. We propose to consider successive recommendations to users under the form of graphs of recommendations. We give models for this representation. Motivated by the growing interest for algorithmic transparency, we then propose a first application for those graphs, that is the potential detection of introduced recommendation bias by the service provider. This application relies on the analysis of the topology of the extracted graph for a given user; we propose a notion of recommendation coherence with regards to the topological proximity of recommended items (under the measure of items' k-closest neighbors, reminding the "small-world" model by Watts & Stroggatz). We finally illustrate this approach on a model and on Youtube crawls, targeting the prediction of "Recommended for you" links (i.e., biased or not by Youtube).

A large portion of data mining and analytic services use modern machine learning techniques, such as deep learning. The state-of-the-art results by deep learning come at the price of an intensive use of computing resources. The leading frameworks (e.g., TensorFlow) are executed on GPUs or on high-end servers in datacenters. On the other end, there is a proliferation of personal devices with possibly free CPU cycles; this can enable services to run in users' homes, embedding machine learning operations. In this paper, we ask the following question: Is distributed deep learning computation on WAN connected devices feasible, in spite of the traffic caused by learning tasks? We show that such a setup rises some important challenges, most notably the ingress traffic that the servers hosting the up-to-date model have to sustain. In order to reduce this stress, we propose adaComp, a novel algorithm for compressing worker updates to the model on the server. Applicable to stochastic gradient descent based approaches, it combines efficient gradient selection and learning rate modulation. We then experiment and measure the impact of compression, device heterogeneity and reliability on the accuracy of learned models, with an emulator platform that embeds TensorFlow into Linux containers. We report a reduction of the total amount of data sent by workers to the server by two order of magnitude (e.g., 191-fold reduction for a convolutional network on the MNIST dataset), when compared to a standard asynchronous stochastic gradient descent, while preserving model accuracy., Comment: Best paper award at IEEE International Symposium on Network Computing and Applications (NCA 2017)

Ensuring the early detection of important social network users is a challenging task. Some peer ranking services are now well established, such as PeerIndex, Klout, or Kred. Their function is to rank users according to their influence. This notion of influence is however abstract, and the algorithms achieving this ranking are opaque. Following the rising demand for a more transparent web, we explore the problem of gaining knowledge by reverse engineering such peer ranking services, with regards to the social network topology they get as an input. Since these services exploit the online activity of users (and therefore their connectivity in social networks), we provide a precise evaluation of how topological metrics of the social network impact the final user ranking. Our approach is the following : we first model the ranking service as a black-box with which we interact by creating user profiles and by performing operations on them. Through those profiles, we trigger some slight topological modifications. By monitoring the impact of these modifications on the rankings of those profiles, we infer the weight of each topological metric in the black-box, thus reversing the service influence cookbook., Comment: To appear in CSCW 2017 (The 20th ACM Conference on Computer-Supported Cooperative Work and Social Computing)

Classical erasure codes, e.g. Reed-Solomon codes, have been acknowledged as an efficient alternative to plain replication to reduce the storage overhead in reliable distributed storage systems. Yet, such codes experience high overhead during the maintenance process. In this paper we propose a novel erasure-coded framework especially tailored for networked storage systems. Our approach relies on the use of random codes coupled with a clustered placement strategy, enabling the maintenance of a failed machine at the granularity of multiple files. Our repair protocol leverages network coding techniques to reduce by half the amount of data transferred during maintenance, as several files can be repaired simultaneously. This approach, as formally proven and demonstrated by our evaluation on a public experimental testbed, enables to dramatically decrease the bandwidth overhead during the maintenance process, as well as the time to repair a failure. In addition, the implementation is made as simple as possible, aiming at a deployment into practical systems., Comment: 14 pages, 13 figures

Group management is a fundamental building block of today's Internet applications. Mailing lists, chat systems, collaborative document edition but also online social networks such as Facebook and Twitter use group management systems. In many cases, group security is required in the sense that access to data is restricted to group members only. Some applications also require privacy by keeping group members anonymous and unlinkable. Group management systems routinely rely on a central authority that manages and controls the infrastructure and data of the system. Personal user data related to groups then becomes de facto accessible to the central authority. In this paper, we propose a completely distributed approach for group management based on distributed hash tables. As there is no enrollment to a central authority, the created groups can be leveraged by various applications. Following this paradigm we describe a protocol for such a system. We consider security and privacy issues inherently introduced by removing the central authority and provide a formal validation of security properties of the system using AVISPA. We demonstrate the feasibility of this protocol by implementing a prototype running on top of Vuze's DHT.

Energy consumption is the most important factor that determines sensor node lifetime. The optimization of wireless sensor network lifetime targets not only the reduction of energy consumption of a single sensor node but also the extension of the entire network lifetime. We propose a simple and adaptive energy-conserving topology management scheme, called SAND (Self-Organizing Active Node Density). SAND is fully decentralized and relies on a distributed probing approach and on the redundancy resolution of sensors for energy optimizations, while preserving the data forwarding and sensing capabilities of the network. We present the SAND's algorithm, its analysis of convergence, and simulation results. Simulation results show that, though slightly increasing path lengths from sensor to sink nodes, the proposed scheme improves significantly the network lifetime for different neighborhood densities degrees, while preserving both sensing and routing fidelity.