Online services often require users to agree to lengthy and obscure Terms of Service (ToS), leading to information asymmetry and legal risks. This paper proposes TOSense—a Chrome extension that allows users to ask questions about ToS in natural language and get concise answers in real time. The system combines (i) a crawler "tos-crawl" that automatically extracts ToS content, and (ii) a lightweight large language model pipeline: MiniLM for semantic retrieval and BART-encoder for answer relevance verification. To avoid expensive manual annotation, we present a novel Question Answering Evaluation Pipeline (QEP) that generates synthetic questions and verifies the correctness of answers using clustered topic matching. Experiments on five major platforms, Apple, Google, X (formerly Twitter), Microsoft, and Netflix, show the effectiveness of TOSense (with up to 44.5% accuracy) across varying number of topic clusters. During the demonstration, we will showcase TOSense in action. Attendees will be able to experience seamless extraction, interactive question answering, and instant indexing of new sites.

Abstract Semantic understanding of roadways is a key enabling factor for safe autonomous driving. However, existing autonomous driving datasets provide well-structured urban roads while ignoring unstructured roadways containing distress, potholes, water puddles, and various kinds of road patches i.e., earthen, gravel etc. To this end, we introduce Road Region Segmentation dataset (R2S100K)—a large-scale dataset and benchmark for training and evaluation of road segmentation in aforementioned challenging unstructured roadways. R2S100K comprises 100K images extracted from a large and diverse set of video sequences covering more than 1000 km of roadways. Out of these 100K privacy respecting images, 14,000 images have fine pixel-labeling of road regions, with 86,000 unlabeled images that can be leveraged through semi-supervised learning methods. Alongside, we present an Efficient Data Sampling based self-training framework to improve learning by leveraging unlabeled data. Our experimental results demonstrate that the proposed method significantly improves learning methods in generalizability and reduces the labeling cost for semantic segmentation tasks. Our benchmark will be publicly available to facilitate future research at https://r2s100k.github.io/ .

To combat the rapid spread of misinformation on social networks, automated misinformation detection systems based on deep neural networks (DNNs) have been developed. However, these tools are often proprietary and lack transparency, which limits their usefulness. Furthermore, privacy concerns limit data sharing by data owners as well as by data-driven misinformation-detection services. Although data encryption techniques can help address privacy concerns in DNN inference, there is a challenge to the seamless integration of these techniques due to the encryption errors induced by cascaded encrypted operations, as well as a mismatch between the tools used for DNNs and cryptography. In this paper, we make two-fold contributions. First, we study the noise bounds of homomorphic encryption (HE) operations as error propagation in DNN layers and derive two properties that, if satisfied by the layer, will considerably reduce the output error. We identify that$L_{2}$regularization and sigmoid activation satisfy these properties and validate our hypothesis, for instance, replacing ReLU with sigmoid reduced the output error by$10^{6}\times$(best case) to$10\times$(worst case). Second, we extend the Python encryption library TenSeal by enabling the automatic conversion of a TensorFlow DNN into an encryption-compatible DNN with a few lines of code. These contributions are significant as encryption-friendly DL architectures are sorely needed to close the gap between DL-in-research and DL-in-practice.

Recent works have shown that Federated Learning (FL) is vulnerable to backdoor attacks. Existing defenses cluster submitted updates from clients and select the best cluster for aggregation. However, they often rely on unrealistic assumptions regarding client submissions and sampled clients population while choosing the best cluster. We show that in realistic FL settings, state-of-the-art (SOTA) defenses struggle to perform well against backdoor attacks in FL. To address this, we highlight that backdoored submissions are adversarially biased and overconfident compared to clean submissions. We, therefore, propose an Adversarially Guided Stateful Defense (AGSD) against backdoor attacks on Deep Neural Networks (DNNs) in FL scenarios. AGSD employs adversarial perturbations to a small held-out dataset to compute a novel metric, called the trust index, that guides the cluster selection without relying on any unrealistic assumptions regarding client submissions. Moreover, AGSD maintains a trust state history of each client that adaptively penalizes backdoored clients and rewards clean clients. In realistic FL settings, where SOTA defenses mostly fail to resist attacks, AGSD mostly outperforms all SOTA defenses with minimal drop in clean accuracy (5% in the worst-case compared to best accuracy) even when (a) given a very small held-out dataset—typically AGSD assumes 50 samples (≤ 0.1% of the training data) and (b) no held-out dataset is available, and out-of-distribution data is used instead. For reproducibility, our code will be openly available at: https://github.com/hassanalikhatim/AGSD.

Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow patterns, which can be used for more effective urban planning and smart city management. However, DL models have been known to perform poorly on inconspicuous adversarial perturbations. Although many works have studied these adversarial perturbations in general, the adversarial vulnerabilities of deep CFP models in particular have remained largely unexplored. In this paper, we perform a rigorous analysis of the adversarial vulnerabilities of DL-based CFP models under multiple threat settings, making three-fold contributions; 1) we propose CaV-detect by formally identifying two novel properties—ConsistencyandValidity—of the CFP inputs that enable thedetection of standard adversarial inputs with 0% false acceptance rate (FAR); 2) we leverage universal adversarial perturbations and an adaptive adversarial loss to present adaptive adversarial attacks to evade CaV-detect defense; 3) we propose CVP, aConsistent,Valid andPhysically-realizable adversarial attack, that explicitly inducts the consistency and validity priors in the perturbation generation mechanism. We find out that although the crowd-flow models are vulnerable to adversarial perturbations, it is extremely challenging to simulate these perturbations in physical settings, notably when CaV-detect is in place. We also show that CVP attack considerably outperforms the adaptively modified standard attacks in FAR and adversarial loss metrics. We conclude with useful insights emerging from our work and highlight promising future research directions.

Deep Learning (DL) algorithms have shown wonders in many Natural Language Processing (NLP) tasks such as language-to-language translation, spam filtering, fake-news detection, and comprehension understanding. However, research has shown that the adversarial vulnerabilities of deep learning networks manifest themselves when DL is used for NLP tasks. Most mitigation techniques proposed to date are supervised—relying on adversarial retraining to improve the robustness—which is impractical. This work introduces a novel, unsupervised detection methodology for detecting adversarial inputs to NLP classifiers. In summary, we note that minimally perturbing an input to change a model’s output—a major strength of adversarial attacks—is a weakness that leaves unique statistical marks reflected in the cumulative contribution scores of the input. Particularly, we show that the cumulative contribution score, called CF-score, of adversarial inputs is generally greater than that of the clean inputs. We thus propose Con-Detect—a Contribution based Detection method—for detecting adversarial attacks against NLP classifiers. Con-Detect can be deployed with any classifier without having to retrain it. We experiment with multiple attackers—Text-bugger, Text-fooler, PWWS—on several architectures—MLP, CNN, LSTM, Hybrid CNN-RNN, BERT—trained for different classification tasks—IMDB sentiment classification, fake-news classification, AG news topic classification—under different threat models—Con-Detect-blind attacks, Con-Detect-aware attacks, and Con-Detect-adaptive attacks—and show that Con-Detect can reduce the attack success rate (ASR) of different attacks from 100% to as low as 0% for the best cases and ≈70% for the worst case. Even in the worst case, we note a 100% increase in the required number of queries and a 50% increase in the number of words perturbed, suggesting that Con-Detect is hard to evade.

The use of artificial intelligence (AI) at the edge is transforming every aspect of the lives of human beings from scheduling daily activities to personalized shopping recommendations. Since the success of AI is to be measured ultimately in terms of how it benefits human beings, and that the data driving the deep learning-based edge AI algorithms are intricately and intimately tied to humans, it is important to look at these AI technologies through a human-centric lens. However, despite the significant impact of AI design on human interests, the security and trustworthiness of edge AI applications are not foolproof and ethicalneither foolproof nor ethical; Moreover, social norms are often ignored duringin the design, implementation, and deployment of edge AI systems. In this paper, we make the following two contributions: Firstly, we analyze the application of edge AI through a human-centric perspective. More specifically, we present a pipeline to develop human-centric embedded machine learning (HC-EML) applications leveraging a generic human-centric AI (HCAI) framework. Alongside, we also analyzediscuss the privacy, trustworthiness, robustness, and security aspects of HC-EML applications with an insider look at their challenges and possible solutions along the way. Secondly, to illustrate the gravity of these issues, we present a case study on the task of human facial emotion recognition (FER) based on AffectNet dataset, where we analyze the effects of widely used input quantization on the security, robustness, fairness, and trustworthiness of an EML model. We find that input quantization partially degrades the efficacy of adversarial and backdoor attacks at the cost of a slight decrease in accuracy over clean inputs. By analyzing the explanations generated by SHAP, we identify that the decision of a FER model is largely influenced by features such as eyes, alar crease, lips, and jaws. Additionally, we note that input quantization is notably biased against the dark skin faces, and hypothesize that low-contrast features of dark skin faces may be responsible for the observed trends. We conclude with precautionary remarks and guidelines for future researchers.

While the technique of Deep Neural Networks (DNNs) has been instrumental in achieving state-of-the-art results for various Natural Language Processing (NLP) tasks, recent works have shown that the decisions made by DNNs cannot always be trusted. Recently Explainable Artificial Intelligence (XAI) methods have been proposed as a method for increasing DNN’s reliability and trustworthiness. These XAI methods are however open to attack and can be manipulated in both white-box (gradient-based) and black-box (perturbation-based) scenarios. Exploring novel techniques to attack and robustify these XAI methods is crucial to fully understand these vulnerabilities. In this work, we propose Tamp-X—a novel attack which tampers the activations of robust NLP classifiers forcing the state-of-the-art white-box and black-box XAI methods to generate misrepresented explanations. To the best of our knowledge, in current NLP literature, we are the first to attack both the white-box and the black-box XAI methods simultaneously. We quantify the reliability of explanations based on three different metrics—the descriptive accuracy, the cosine similarity, and the Lp norms of the explanation vectors. Through extensive experimentation, we show that the explanations generated for the tampered classifiers are not reliable, and significantly disagree with those generated for the untampered classifiers despite that the output decisions of tampered and untampered classifiers are almost always the same. Additionally, we study the adversarial robustness of the tampered NLP classifiers, and find out that the tampered classifiers which are harder to explain for the XAI methods, are also harder to attack by the adversarial attackers.

Due to the excessive use of cloud-based machine learning (ML) services, the smart cyber-physical systems (CPS) are increasingly becoming vulnerable to black-box attacks on their ML modules. Traditionally, the black-box attacks are either transfer attacks requiring model stealing, or score/decision-based gradient estimation attacks requiring a large number of queries. In practical scenarios, especially for cloud-based ML services and timing-constrained CPS use-cases, every query incurs a huge cost, thereby rendering state-of-the-art decision-based attacks ineffective in such settings. Towards this, we propose a novel methodology for automatically generating an extremely fast and imperceptible decision-based attack called FaDec. It follows two main steps: (1) fast estimation of the classification boundary by combining the half-interval search-based algorithm with gradient sign estimation to reduce the number of queries; and (2) adversarial noise optimization to ensure the imperceptibility. For illustration, we evaluate FaDec on the image recognition and traffic sign detection using multiple state-of-the-art DNNs trained on CIFAR-10 and the German Traffic Sign Recognition Benchmarks (GTSRB) datasets. The experimental analysis shows that the proposed FaDec attack is 16x faster compared to the state-of-the-art decision-based attacks, and generates an attack image with better imperceptibility for a much lesser number of iterations, thereby making our attack more powerful in practical scenarios. We open-sourced the complete code and results of our methodology at https://github.com/fklodhi/FaDec.

Adversarial examples have emerged as a significant threat to machine learning algorithms, especially to the convolutional neural networks (CNNs). In this paper, we propose two quantization-based defense mechanisms, Constant Quantization (CQ) and Trainable Quantization (TQ), to increase the robustness of CNNs against adversarial examples. CQ quantizes input pixel intensities based on a “fixed” number of quantization levels, while in TQ, the quantization levels are “iteratively learned during the training phase”, thereby providing a stronger defense mechanism. We apply the proposed techniques on undefended CNNs against different state-of-the-art adversarial attacks from the open-source Cleverhans library. The experimental results demonstrate 50%-96% and 10%-50% increase in the classification accuracy of the perturbed images generated from the MNIST and the CIFAR-10 datasets, respectively, on commonly used CNN (Conv2D(64, 8×8)-Conv2D(128, 6×6)-Conv2D(128, 5×5) - Dense(10) - Softmax()) available in Cleverhans library.