What's Pulling the Strings? Evaluating Integrity and Attribution in AI Training and Inference through Concept Shift

The growing adoption of artificial intelligence (AI) has amplified concerns about trustworthiness, including integrity, privacy, robustness, and bias. To assess and attribute these threats, we propose ConceptLens, a generic framework that leverages pre-trained multimodal models to identify the root causes of integrity threats by analyzing Concept Shift in probing samples. ConceptLens demonstrates strong detection performance for vanilla data poisoning attacks and uncovers vulnerabilities to bias injection, such as the generation of covert advertisements through malicious concept shifts. It identifies privacy risks in unaltered but high-risk samples, filters them before training, and provides insights into model weaknesses arising from incomplete or imbalanced training data. Additionally, at the model level, it attributes concepts that the target model is overly dependent on, identifies misleading concepts, and explains how disrupting key concepts negatively impacts the model. Furthermore, it uncovers sociological biases in generative content, revealing disparities across sociological contexts. Strikingly, ConceptLens reveals how safe training and inference data can be unintentionally and easily exploited, potentially undermining safety alignment. Our study informs actionable insights to breed trust in AI systems, thereby speeding adoption and driving greater innovation.

Whispering Under the Eaves: Protecting User Privacy Against Commercial and LLM-powered Automatic Speech Recognition Systems

The widespread application of automatic speech recognition (ASR) supports large-scale voice surveillance, raising concerns about privacy among users. In this paper, we concentrate on using adversarial examples to mitigate unauthorized disclosure of speech privacy thwarted by potential eavesdroppers in speech communications. While audio adversarial examples have demonstrated the capability to mislead ASR models or evade ASR surveillance, they are typically constructed through time-intensive offline optimization, restricting their practicality in real-time voice communication. Recent work overcame this limitation by generating universal adversarial perturbations (UAPs) and enhancing their transferability for black-box scenarios. However, they introduced excessive noise that significantly degrades audio quality and affects human perception, thereby limiting their effectiveness in practical scenarios. To address this limitation and protect live users' speech against ASR systems, we propose a novel framework, AudioShield. Central to this framework is the concept of Transferable Universal Adversarial Perturbations in the Latent Space (LS-TUAP). By transferring the perturbations to the latent space, the audio quality is preserved to a large extent. Additionally, we propose target feature adaptation to enhance the transferability of UAPs by embedding target text features into the perturbations. Comprehensive evaluation on four commercial ASR APIs (Google, Amazon, iFlytek, and Alibaba), three voice assistants, two LLM-powered ASR and one NN-based ASR demonstrates the protection superiority of AudioShield over existing competitors, and both objective and subjective evaluations indicate that AudioShield significantly improves the audio quality. Moreover, AudioShield also shows high effectiveness in real-time end-to-end scenarios, and demonstrates strong resilience against adaptive countermeasures.

From Pixels to Trajectory: Universal Adversarial Example Detection via Temporal Imprints

For the first time, we unveil discernible temporal (or historical) trajectory imprints resulting from adversarial example (AE) attacks. Standing in contrast to existing studies all focusing on spatial (or static) imprints within the targeted underlying victim models, we present a fresh temporal paradigm for understanding these attacks. Of paramount discovery is that these imprints are encapsulated within a single loss metric, spanning universally across diverse tasks such as classification and regression, and modalities including image, text, and audio. Recognizing the distinct nature of loss between adversarial and clean examples, we exploit this temporal imprint for AE detection by proposing TRAIT (TRaceable Adversarial temporal trajectory ImprinTs). TRAIT operates under minimal assumptions without prior knowledge of attacks, thereby framing the detection challenge as a one-class classification problem. However, detecting AEs is still challenged by significant overlaps between the constructed synthetic losses of adversarial and clean examples due to the absence of ground truth for incoming inputs. TRAIT addresses this challenge by converting the synthetic loss into a spectrum signature, using the technique of Fast Fourier Transform to highlight the discrepancies, drawing inspiration from the temporal nature of the imprints, analogous to time-series signals. Across 12 AE attacks including SMACK (USENIX Sec'2023), TRAIT demonstrates consistent outstanding performance across comprehensively evaluated modalities, tasks, datasets, and model architectures. In all scenarios, TRAIT achieves an AE detection accuracy exceeding 97%, often around 99%, while maintaining a false rejection rate of 1%. TRAIT remains effective under the formulated strong adaptive attacks.

CAMP in the Odyssey: Provably Robust Reinforcement Learning with Certified Radius Maximization

Deep reinforcement learning (DRL) has gained widespread adoption in control and decision-making tasks due to its strong performance in dynamic environments. However, DRL agents are vulnerable to noisy observations and adversarial attacks, and concerns about the adversarial robustness of DRL systems have emerged. Recent efforts have focused on addressing these robustness issues by establishing rigorous theoretical guarantees for the returns achieved by DRL agents in adversarial settings. Among these approaches, policy smoothing has proven to be an effective and scalable method for certifying the robustness of DRL agents. Nevertheless, existing certifiably robust DRL relies on policies trained with simple Gaussian augmentations, resulting in a suboptimal trade-off between certified robustness and certified return. To address this issue, we introduce a novel paradigm dubbed \texttt{C}ertified-r\texttt{A}dius-\texttt{M}aximizing \texttt{P}olicy (\texttt{CAMP}) training. \texttt{CAMP} is designed to enhance DRL policies, achieving better utility without compromising provable robustness. By leveraging the insight that the global certified radius can be derived from local certified radii based on training-time statistics, \texttt{CAMP} formulates a surrogate loss related to the local certified radius and optimizes the policy guided by this surrogate loss. We also introduce \textit{policy imitation} as a novel technique to stabilize \texttt{CAMP} training. Experimental results demonstrate that \texttt{CAMP} significantly improves the robustness-return trade-off across various tasks. Based on the results, \texttt{CAMP} can achieve up to twice the certified expected return compared to that of baselines. Our code is available at https://github.com/NeuralSec/camp-robust-rl.

AI-Compass: A Comprehensive and Effective Multi-module Testing Tool for AI Systems

AI systems, in particular with deep learning techniques, have demonstrated superior performance for various real-world applications. Given the need for tailored optimization in specific scenarios, as well as the concerns related to the exploits of subsurface vulnerabilities, a more comprehensive and in-depth testing AI system becomes a pivotal topic. We have seen the emergence of testing tools in real-world applications that aim to expand testing capabilities. However, they often concentrate on ad-hoc tasks, rendering them unsuitable for simultaneously testing multiple aspects or components. Furthermore, trustworthiness issues arising from adversarial attacks and the challenge of interpreting deep learning models pose new challenges for developing more comprehensive and in-depth AI system testing tools. In this study, we design and implement a testing tool, \tool, to comprehensively and effectively evaluate AI systems. The tool extensively assesses multiple measurements towards adversarial robustness, model interpretability, and performs neuron analysis. The feasibility of the proposed testing tool is thoroughly validated across various modalities, including image classification, object detection, and text classification. Extensive experiments demonstrate that \tool is the state-of-the-art tool for a comprehensive assessment of the robustness and trustworthiness of AI systems. Our research sheds light on a general solution for AI systems testing landscape.

Reconstruction of Differentially Private Text Sanitization via Large Language Models

Differential privacy (DP) is the de facto privacy standard against privacy leakage attacks, including many recently discovered ones against large language models (LLMs). However, we discovered that LLMs could reconstruct the altered/removed privacy from given DP-sanitized prompts. We propose two attacks (black-box and white-box) based on the accessibility to LLMs and show that LLMs could connect the pair of DP-sanitized text and the corresponding private training data of LLMs by giving sample text pairs as instructions (in the black-box attacks) or fine-tuning data (in the white-box attacks). To illustrate our findings, we conduct comprehensive experiments on modern LLMs (e.g., LLaMA-2, LLaMA-3, ChatGPT-3.5, ChatGPT-4, ChatGPT-4o, Claude-3, Claude-3.5, OPT, GPT-Neo, GPT-J, Gemma-2, and Pythia) using commonly used datasets (such as WikiMIA, Pile-CC, and Pile-Wiki) against both word-level and sentence-level DP. The experimental results show promising recovery rates, e.g., the black-box attacks against the word-level DP over WikiMIA dataset gave 72.18% on LLaMA-2 (70B), 82.39% on LLaMA-3 (70B), 75.35% on Gemma-2, 91.2% on ChatGPT-4o, and 94.01% on Claude-3.5 (Sonnet). More urgently, this study indicates that these well-known LLMs have emerged as a new security risk for existing DP text sanitization approaches in the current environment.

Edge Unlearning is Not "on Edge"! An Adaptive Exact Unlearning System on Resource-Constrained Devices

The right to be forgotten mandates that machine learning models enable the erasure of a data owner's data and information from a trained model. Removing data from the dataset alone is inadequate, as machine learning models can memorize information from the training data, increasing the potential privacy risk to users. To address this, multiple machine unlearning techniques have been developed and deployed. Among them, approximate unlearning is a popular solution, but recent studies report that its unlearning effectiveness is not fully guaranteed. Another approach, exact unlearning, tackles this issue by discarding the data and retraining the model from scratch, but at the cost of considerable computational and memory resources. However, not all devices have the capability to perform such retraining. In numerous machine learning applications, such as edge devices, Internet-of-Things (IoT), mobile devices, and satellites, resources are constrained, posing challenges for deploying existing exact unlearning methods. In this study, we propose a Constraint-aware Adaptive Exact Unlearning System at the network Edge (CAUSE), an approach to enabling exact unlearning on resource-constrained devices. Aiming to minimize the retrain overhead by storing sub-models on the resource-constrained device, CAUSE innovatively applies a Fibonacci-based replacement strategy and updates the number of shards adaptively in the user-based data partition process. To further improve the effectiveness of memory usage, CAUSE leverages the advantage of model pruning to save memory via compression with minimal accuracy sacrifice. The experimental results demonstrate that CAUSE significantly outperforms other representative systems in realizing exact unlearning on the resource-constrained device by 9.23%-80.86%, 66.21%-83.46%, and 5.26%-194.13% in terms of unlearning speed, energy consumption, and accuracy.

Rethinking the Threat and Accessibility of Adversarial Attacks against Face Recognition Systems

Face recognition pipelines have been widely deployed in various mission-critical systems in trust, equitable and responsible AI applications. However, the emergence of adversarial attacks has threatened the security of the entire recognition pipeline. Despite the sheer number of attack methods proposed for crafting adversarial examples in both digital and physical forms, it is never an easy task to assess the real threat level of different attacks and obtain useful insight into the key risks confronted by face recognition systems. Traditional attacks view imperceptibility as the most important measurement to keep perturbations stealthy, while we suspect that industry professionals may possess a different opinion. In this paper, we delve into measuring the threat brought about by adversarial attacks from the perspectives of the industry and the applications of face recognition. In contrast to widely studied sophisticated attacks in the field, we propose an effective yet easy-to-launch physical adversarial attack, named AdvColor, against black-box face recognition pipelines in the physical world. AdvColor fools models in the recognition pipeline via directly supplying printed photos of human faces to the system under adversarial illuminations. Experimental results show that physical AdvColor examples can achieve a fooling rate of more than 96% against the anti-spoofing model and an overall attack success rate of 88% against the face recognition pipeline. We also conduct a survey on the threats of prevailing adversarial attacks, including AdvColor, to understand the gap between the machine-measured and human-assessed threat levels of different forms of adversarial attacks. The survey results surprisingly indicate that, compared to deliberately launched imperceptible attacks, perceptible but accessible attacks pose more lethal threats to real-world commercial systems of face recognition.

QUEEN: Query Unlearning against Model Extraction

Model extraction attacks currently pose a non-negligible threat to the security and privacy of deep learning models. By querying the model with a small dataset and usingthe query results as the ground-truth labels, an adversary can steal a piracy model with performance comparable to the original model. Two key issues that cause the threat are, on the one hand, accurate and unlimited queries can be obtained by the adversary; on the other hand, the adversary can aggregate the query results to train the model step by step. The existing defenses usually employ model watermarking or fingerprinting to protect the ownership. However, these methods cannot proactively prevent the violation from happening. To mitigate the threat, we propose QUEEN (QUEry unlEarNing) that proactively launches counterattacks on potential model extraction attacks from the very beginning. To limit the potential threat, QUEEN has sensitivity measurement and outputs perturbation that prevents the adversary from training a piracy model with high performance. In sensitivity measurement, QUEEN measures the single query sensitivity by its distance from the center of its cluster in the feature space. To reduce the learning accuracy of attacks, for the highly sensitive query batch, QUEEN applies query unlearning, which is implemented by gradient reverse to perturb the softmax output such that the piracy model will generate reverse gradients to worsen its performance unconsciously. Experiments show that QUEEN outperforms the state-of-the-art defenses against various model extraction attacks with a relatively low cost to the model accuracy. The artifact is publicly available at https://anonymous.4open.science/r/queen implementation-5408/.

On Security Weaknesses and Vulnerabilities in Deep Learning Systems

The security guarantee of AI-enabled software systems (particularly using deep learning techniques as a functional core) is pivotal against the adversarial attacks exploiting software vulnerabilities. However, little attention has been paid to a systematic investigation of vulnerabilities in such systems. A common situation learned from the open source software community is that deep learning engineers frequently integrate off-the-shelf or open-source learning frameworks into their ecosystems. In this work, we specifically look into deep learning (DL) framework and perform the first systematic study of vulnerabilities in DL systems through a comprehensive analysis of identified vulnerabilities from Common Vulnerabilities and Exposures (CVE) and open-source DL tools, including TensorFlow, Caffe, OpenCV, Keras, and PyTorch. We propose a two-stream data analysis framework to explore vulnerability patterns from various databases. We investigate the unique DL frameworks and libraries development ecosystems that appear to be decentralized and fragmented. By revisiting the Common Weakness Enumeration (CWE) List, which provides the traditional software vulnerability related practices, we observed that it is more challenging to detect and fix the vulnerabilities throughout the DL systems lifecycle. Moreover, we conducted a large-scale empirical study of 3,049 DL vulnerabilities to better understand the patterns of vulnerability and the challenges in fixing them. We have released the full replication package at https://github.com/codelzz/Vulnerabilities4DLSystem. We anticipate that our study can advance the development of secure DL systems.