This paper investigates the weaknesses of image watermarking techniques. We present WAVES (Watermark Analysis Via Enhanced Stress-testing), a novel benchmark for assessing watermark robustness, overcoming the limitations of current evaluation methods.WAVES integrates detection and identification tasks, and establishes a standardized evaluation protocol comprised of a diverse range of stress tests. The attacks in WAVES range from traditional image distortions to advanced and novel variations of diffusive, and adversarial attacks. Our evaluation examines two pivotal dimensions: the degree of image quality degradation and the efficacy of watermark detection after attacks. We develop a series of Performance vs. Quality 2D plots, varying over several prominent image similarity metrics, which are then aggregated in a heuristically novel manner to paint an overall picture of watermark robustness and attack potency. Our comprehensive evaluation reveals previously undetected vulnerabilities of several modern watermarking algorithms. We envision WAVES as a toolkit for the future development of robust watermarking systems. The project is available at https://wavesbench.github.io/
Safety alignment of Large Language Models (LLMs) can be compromised with manual jailbreak attacks and (automatic) adversarial attacks. Recent work suggests that patching LLMs against these attacks is possible: manual jailbreak attacks are human-readable but often limited and public, making them easy to block; adversarial attacks generate gibberish prompts that can be detected using perplexity-based filters. In this paper, we show that these solutions may be too optimistic. We propose an interpretable adversarial attack, \texttt{AutoDAN}, that combines the strengths of both types of attacks. It automatically generates attack prompts that bypass perplexity-based filters while maintaining a high attack success rate like manual jailbreak attacks. These prompts are interpretable and diverse, exhibiting strategies commonly used in manual jailbreak attacks, and transfer better than their non-readable counterparts when using limited training data or a single proxy model. We also customize \texttt{AutoDAN}'s objective to leak system prompts, another jailbreak application not addressed in the adversarial attack literature. %, demonstrating the versatility of the approach. We can also customize the objective of \texttt{AutoDAN} to leak system prompts, beyond the ability to elicit harmful content from the model, demonstrating the versatility of the approach. Our work provides a new way to red-team LLMs and to understand the mechanism of jailbreak attacks.
CLIP, as a foundational vision language model, is widely used in zero-shot image classification due to its ability to understand various visual concepts and natural language descriptions. However, how to fully leverage CLIP's unprecedented human-like understanding capabilities to achieve better zero-shot classification is still an open question. This paper draws inspiration from the human visual perception process: a modern neuroscience view suggests that in classifying an object, humans first infer its class-independent attributes (e.g., background and orientation) which help separate the foreground object from the background, and then make decisions based on this information. Inspired by this, we observe that providing CLIP with contextual attributes improves zero-shot classification and mitigates reliance on spurious features. We also observe that CLIP itself can reasonably infer the attributes from an image. With these observations, we propose a training-free, two-step zero-shot classification method named PerceptionCLIP. Given an image, it first infers contextual attributes (e.g., background) and then performs object classification conditioning on them. Our experiments show that PerceptionCLIP achieves better generalization, group robustness, and better interpretability. For example, PerceptionCLIP with ViT-L/14 improves the worst group accuracy by 16.5% on the Waterbirds dataset and by 3.5% on CelebA.
Our work focuses on the challenge of detecting outputs generated by Large Language Models (LLMs) from those generated by humans. The ability to distinguish between the two is of utmost importance in numerous applications. However, the possibility and impossibility of such discernment have been subjects of debate within the community. Therefore, a central question is whether we can detect AI-generated text and, if so, when. In this work, we provide evidence that it should almost always be possible to detect the AI-generated text unless the distributions of human and machine generated texts are exactly the same over the entire support. This observation follows from the standard results in information theory and relies on the fact that if the machine text is becoming more like a human, we need more samples to detect it. We derive a precise sample complexity bound of AI-generated text detection, which tells how many samples are needed to detect. This gives rise to additional challenges of designing more complicated detectors that take in n samples to detect than just one, which is the scope of future research on this topic. Our empirical evaluations support our claim about the existence of better detectors demonstrating that AI-Generated text detection should be achievable in the majority of scenarios. Our results emphasize the importance of continued research in this area
Machine learning models that are developed to be invariant under certain types of data transformations have shown improved generalization in practice. However, a principled understanding of why invariance benefits generalization is limited. Given a dataset, there is often no principled way to select "suitable" data transformations under which model invariance guarantees better generalization. This paper studies the generalization benefit of model invariance by introducing the sample cover induced by transformations, i.e., a representative subset of a dataset that can approximately recover the whole dataset using transformations. For any data transformations, we provide refined generalization bounds for invariant models based on the sample cover. We also characterize the "suitability" of a set of data transformations by the sample covering number induced by transformations, i.e., the smallest size of its induced sample covers. We show that we may tighten the generalization bounds for "suitable" transformations that have a small sample covering number. In addition, our proposed sample covering number can be empirically evaluated and thus provides a guide for selecting transformations to develop model invariance for better generalization. In experiments on multiple datasets, we evaluate sample covering numbers for some commonly used transformations and show that the smaller sample covering number for a set of transformations (e.g., the 3D-view transformation) indicates a smaller gap between the test and training error for invariant models, which verifies our propositions.
Training machine learning models to be robust against adversarial inputs poses seemingly insurmountable challenges. To better understand model robustness, we consider the underlying problem of learning robust representations. We develop a general definition of representation vulnerability that captures the maximum change of mutual information between the input and output distributions, under the worst-case input distribution perturbation. We prove a theorem that establishes a lower bound on the minimum adversarial risk that can be achieved for any downstream classifier based on this definition. We then propose an unsupervised learning method for obtaining intrinsically robust representations by maximizing the worst-case mutual information between input and output distributions. Experiments on downstream classification tasks and analyses of saliency maps support the robustness of the representations found using unsupervised learning with our training principle.
Machine learning models are generally vulnerable to adversarial examples, which is in contrast to the robustness of humans. In this paper, we try to leverage one of the mechanisms in human recognition and propose a bio-inspired classification framework in which model inference is conditioned on label hypothesis. We provide a class of training objectives for this framework and an information bottleneck regularizer which utilizes the advantage that label information can be discarded during inference. This framework enables better compression of the mutual information between inputs and latent representations without loss of learning capacity, at the cost of tractable inference complexity. Better compression and elimination of label information further bring better adversarial robustness without loss of natural accuracy, which is demonstrated in the experiment.