With the widespread deployment of deep neural network (DNN) models, dynamic watermarking techniques are being used to protect the intellectual property of model owners. However, recent studies have shown that existing watermarking schemes are vulnerable to watermark removal and ambiguity attacks. Besides, the vague criteria for determining watermark presence further increase the likelihood of such attacks. In this paper, we propose a secure DNN watermarking scheme named ChainMarks, which generates secure and robust watermarks by introducing a cryptographic chain into the trigger inputs and utilizes a two-phase Monte Carlo method for determining watermark presence. First, ChainMarks generates trigger inputs as a watermark dataset by repeatedly applying a hash function over a secret key, where the target labels associated with trigger inputs are generated from the digital signature of model owner. Then, the watermarked model is produced by training a DNN over both the original and watermark datasets. To verify watermarks, we compare the predicted labels of trigger inputs with the target labels and determine ownership with a more accurate decision threshold that considers the classification probability of specific models. Experimental results show that ChainMarks exhibits higher levels of robustness and security compared to state-of-the-art watermarking schemes. With a better marginal utility, ChainMarks provides a higher probability guarantee of watermark presence in DNN models with the same level of watermark accuracy.

As a new paradigm in machine learning, self-supervised learning (SSL) is capable of learning high-quality representations of complex data without relying on labels. In addition to eliminating the need for labeled data, research has found that SSL improves the adversarial robustness over supervised learning since lacking labels makes it more challenging for adversaries to manipulate model predictions. However, the extent to which this robustness superiority generalizes to other types of attacks remains an open question. We explore this question in the context of backdoor attacks. Specifically, we design and evaluate CTRL, an embarrassingly simple yet highly effective self-supervised backdoor attack. By only polluting a tiny fraction of training data (<= 1%) with indistinguishable poisoning samples, CTRL causes any trigger-embedded input to be misclassified to the adversary's designated class with a high probability (>= 99%) at inference time. Our findings suggest that SSL and supervised learning are comparably vulnerable to backdoor attacks. More importantly, through the lens of CTRL, we study the inherent vulnerability of SSL to backdoor attacks. With both empirical and analytical evidence, we reveal that the representation invariance property of SSL, which benefits adversarial robustness, may also be the very reason making \ssl highly susceptible to backdoor attacks. Our findings also imply that the existing defenses against supervised backdoor attacks are not easily retrofitted to the unique vulnerability of SSL.

Logic locking has been proposed to safeguard intellectual property (IP) during chip fabrication. Logic locking techniques protect hardware IP by making a subset of combinational modules in a design dependent on a secret key that is withheld from untrusted parties. If an incorrect secret key is used, a set of deterministic errors is produced in locked modules, restricting unauthorized use. A common target for logic locking is neural accelerators, especially as machine-learning-as-a-service becomes more prevalent. In this work, we explore how logic locking can be used to compromise the security of a neural accelerator it protects. Specifically, we show how the deterministic errors caused by incorrect keys can be harnessed to produce neural-trojan-style backdoors. To do so, we first outline a motivational attack scenario where a carefully chosen incorrect key, which we call a trojan key, produces misclassifications for an attacker-specified input class in a locked accelerator. We then develop a theoretically-robust attack methodology to automatically identify trojan keys. To evaluate this attack, we launch it on several locked accelerators. In our largest benchmark accelerator, our attack identified a trojan key that caused a 74\% decrease in classification accuracy for attacker-specified trigger inputs, while degrading accuracy by only 1.7\% for other inputs on average.

Neural backdoors represent one primary threat to the security of deep learning systems. The intensive research has produced a plethora of backdoor attacks/defenses, resulting in a constant arms race. However, due to the lack of evaluation benchmarks, many critical questions remain under-explored: (i) what are the strengths and limitations of different attacks/defenses? (ii) what are the best practices to operate them? and (iii) how can the existing attacks/defenses be further improved? To bridge this gap, we design and implement TROJANZOO, the first open-source platform for evaluating neural backdoor attacks/defenses in a unified, holistic, and practical manner. Thus far, focusing on the computer vision domain, it has incorporated 8 representative attacks, 14 state-of-the-art defenses, 6 attack performance metrics, 10 defense utility metrics, as well as rich tools for in-depth analysis of the attack-defense interactions. Leveraging TROJANZOO, we conduct a systematic study on the existing attacks/defenses, unveiling their complex design spectrum: both manifest intricate trade-offs among multiple desiderata (e.g., the effectiveness, evasiveness, and transferability of attacks). We further explore improving the existing attacks/defenses, leading to a number of interesting findings: (i) one-pixel triggers often suffice; (ii) training from scratch often outperforms perturbing benign models to craft trojan models; (iii) optimizing triggers and trojan models jointly greatly improves both attack effectiveness and evasiveness; (iv) individual defenses can often be evaded by adaptive attacks; and (v) exploiting model interpretability significantly improves defense robustness. We envision that TROJANZOO will serve as a valuable platform to facilitate future research on neural backdoors.

A backdoor deep learning (DL) model behaves normally upon clean inputs but misbehaves upon trigger inputs as the backdoor attacker desires, posing severe consequences to DL model deployments. State-of-the-art defenses are either limited to specific backdoor attacks (source-agnostic attacks) or non-user-friendly in that machine learning (ML) expertise or expensive computing resources are required. This work observes that all existing backdoor attacks have an inevitable intrinsic weakness, non-transferability, that is, a trigger input hijacks a backdoored model but cannot be effective to another model that has not been implanted with the same backdoor. With this key observation, we propose non-transferability enabled backdoor detection (NTD) to identify trigger inputs for a model-under-test (MUT) during run-time.Specifically, NTD allows a potentially backdoored MUT to predict a class for an input. In the meantime, NTD leverages a feature extractor (FE) to extract feature vectors for the input and a group of samples randomly picked from its predicted class, and then compares similarity between the input and the samples in the FE's latent space. If the similarity is low, the input is an adversarial trigger input; otherwise, benign. The FE is a free pre-trained model privately reserved from open platforms. As the FE and MUT are from different sources, the attacker is very unlikely to insert the same backdoor into both of them. Because of non-transferability, a trigger effect that does work on the MUT cannot be transferred to the FE, making NTD effective against different types of backdoor attacks. We evaluate NTD on three popular customized tasks such as face recognition, traffic sign recognition and general animal classification, results of which affirm that NDT has high effectiveness (low false acceptance rate) and usability (low false rejection rate) with low detection latency.

There is currently a burgeoning demand for deploying deep learning (DL) models on ubiquitous edge Internet of Things devices attributing to their low latency and high privacy preservation. However, DL models are often large in size and require large-scale computation, which prevents them from being placed directly onto IoT devices where resources are constrained and 32-bit floating-point operations are unavailable. Model quantization is a pragmatic solution, which enables DL deployment on mobile devices and embedded systems by effortlessly post-quantizing a large high-precision model into a small low-precision model while retaining the model inference accuracy. This work reveals that the standard quantization operation can be abused to activate a backdoor. We demonstrate that a full-precision backdoored model that does not have any backdoor effect in the presence of a trigger -- as the backdoor is dormant -- can be activated by the default TensorFlow-Lite quantization, the only product-ready quantization framework to date. We ascertain that all trained float-32 backdoored models exhibit no backdoor effect even in the presence of trigger inputs. State-of-the-art frontend detection approaches, such as Neural Cleanse and STRIP, fail to identify the backdoor in the float-32 models. When each of the float-32 models is converted into an int-8 format model through the standard TFLite post-training quantization, the backdoor is activated in the quantized model, which shows a stable attack success rate close to 100% upon inputs with the trigger, while behaves normally upon non-trigger inputs. This work highlights that a stealthy security threat occurs when end users utilize the on-device post-training model quantization toolkits, informing security researchers of cross-platform overhaul of DL models post quantization even if they pass frontend inspections.

Ever since Machine Learning as a Service (MLaaS) emerges as a viable business that utilizes deep learning models to generate lucrative revenue, Intellectual Property Right (IPR) has become a major concern because these deep learning models can easily be replicated, shared, and re-distributed by any unauthorized third parties. To the best of our knowledge, one of the prominent deep learning models - Generative Adversarial Networks (GANs) which has been widely used to create photorealistic image are totally unprotected despite the existence of pioneering IPR protection methodology for Convolutional Neural Networks (CNNs). This paper therefore presents a complete protection framework in both black-box and white-box settings to enforce IPR protection on GANs. Empirically, we show that the proposed method does not compromise the original GANs performance (i.e. image generation, image super-resolution, style transfer), and at the same time, it is able to withstand both removal and ambiguity attacks against embedded watermarks.

In this paper a neuro-robotics model capable of counting using gestures is introduced. The contribution of gestures to learning to count is tested with various model and training conditions. Two studies were presented in this article. In the first, we combine different modalities of the robot's neural network, in the second, a novel training procedure for it is proposed. The model is trained with pointing data from an iCub robot simulator. The behaviour of the model is in line with that of human children in terms of performance change depending on gesture production.

In this work, we attempt to answer a critical question: whether there exists some input sequence that will cause a well-trained discrete-space neural network sequence-to-sequence (seq2seq) model to generate egregious outputs (aggressive, malicious, attacking, etc.). And if such inputs exist, how to find them efficiently. We adopt an empirical methodology, in which we first create lists of egregious output sequences, and then design a discrete optimization algorithm to find input sequences that will cause the model to generate them. Moreover, the optimization algorithm is enhanced for large vocabulary search and constrained to search for input sequences that are likely to be input by real-world users. In our experiments, we apply this approach to dialogue response generation models trained on three real-world dialogue data-sets: Ubuntu, Switchboard and OpenSubtitles, testing whether the model can generate malicious responses. We demonstrate that given the trigger inputs our algorithm finds, a significant number of malicious sentences are assigned large probability by the model, which reveals an undesirable consequence of standard seq2seq training. Recently, research on adversarial attacks (Goodfellow et al., 2014; Szegedy et al., 2013) has been gaining increasing attention: it has been found that for trained deep neural networks (DNNs), when an imperceptible perturbation is applied to the input, the output of the model can change significantly (from correct to incorrect). This line of research has serious implications for our understanding of deep learning models and how we can apply them securely in real-world applications. It has also motivated researchers to design new models or training procedures (Madry et al., 2017), to make the model more robust to those attacks.