Tudor Dumitraş

Sanghyun Hong, Michael-Andrei Panaitescu-Liess, Yiğitcan Kaya et al.

Quantization is a popular technique that $transforms$ the parameter representation of a neural network from floating-point numbers into lower-precision ones ($e.g.$, 8-bit integers). It reduces the memory footprint and the computational cost at inference, facilitating the deployment of resource-hungry models. However, the parameter perturbations caused by this transformation result in $behavioral$ $disparities$ between the model before and after quantization. For example, a quantized model can misclassify some test-time samples that are otherwise classified correctly. It is not known whether such differences lead to a new security vulnerability. We hypothesize that an adversary may control this disparity to introduce specific behaviors that activate upon quantization. To study this hypothesis, we weaponize quantization-aware training and propose a new training framework to implement adversarial quantization outcomes. Following this framework, we present three attacks we carry out with quantization: (i) an indiscriminate attack for significant accuracy loss; (ii) a targeted attack against specific samples; and (iii) a backdoor attack for controlling the model with an input trigger. We further show that a single compromised model defeats multiple quantization schemes, including robust quantization techniques. Moreover, in a federated learning scenario, we demonstrate that a set of malicious participants who conspire can inject our quantization-activated backdoor. Lastly, we discuss potential counter-measures and show that only re-training consistently removes the attack artifacts. Our code is available at https://github.com/Secure-AI-Systems-Group/Qu-ANTI-zation

Sanghyun Hong, Yiğitcan Kaya, Ionuţ-Vlad Modoranu et al.

Recent increases in the computational demands of deep neural networks (DNNs), combined with the observation that most input samples require only simple models, have sparked interest in $input$-$adaptive$ multi-exit architectures, such as MSDNets or Shallow-Deep Networks. These architectures enable faster inferences and could bring DNNs to low-power devices, e.g., in the Internet of Things (IoT). However, it is unknown if the computational savings provided by this approach are robust against adversarial pressure. In particular, an adversary may aim to slowdown adaptive DNNs by increasing their average inference time$-$a threat analogous to the $denial$-$of$-$service$ attacks from the Internet. In this paper, we conduct a systematic evaluation of this threat by experimenting with three generic multi-exit DNNs (based on VGG16, MobileNet, and ResNet56) and a custom multi-exit architecture, on two popular image classification benchmarks (CIFAR-10 and Tiny ImageNet). To this end, we show that adversarial example-crafting techniques can be modified to cause slowdown, and we propose a metric for comparing their impact on different architectures. We show that a slowdown attack reduces the efficacy of multi-exit DNNs by 90-100%, and it amplifies the latency by 1.5-5$\times$ in a typical IoT deployment. We also show that it is possible to craft universal, reusable perturbations and that the attack can be effective in realistic black-box scenarios, where the attacker has limited knowledge about the victim. Finally, we show that adversarial training provides limited protection against slowdowns. These results suggest that further research is needed for defending multi-exit architectures against this emerging threat. Our code is available at https://github.com/sanghyun-hong/deepsloth.

Michael-Andrei Panaitescu-Liess, Yigitcan Kaya, Sicheng Zhu et al.

Group robustness has become a major concern in machine learning (ML) as conventional training paradigms were found to produce high error on minority groups. Without explicit group annotations, proposed solutions rely on heuristics that aim to identify and then amplify the minority samples during training. In our work, we first uncover a critical shortcoming of these methods: an inability to distinguish legitimate minority samples from poison samples in the training set. By amplifying poison samples as well, group robustness methods inadvertently boost the success rate of an adversary -- e.g., from $0\%$ without amplification to over $97\%$ with it. Notably, we supplement our empirical evidence with an impossibility result proving this inability of a standard heuristic under some assumptions. Moreover, scrutinizing recent poisoning defenses both in centralized and federated learning, we observe that they rely on similar heuristics to identify which samples should be eliminated as poisons. In consequence, minority samples are eliminated along with poisons, which damages group robustness -- e.g., from $55\%$ without the removal of the minority samples to $41\%$ with it. Finally, as they pursue opposing goals using similar heuristics, our attempt to alleviate the trade-off by combining group robustness methods and poisoning defenses falls short. By exposing this tension, we also hope to highlight how benchmark-driven ML scholarship can obscure the trade-offs among different metrics with potentially detrimental consequences.

Octavian Suciu, Connor Nelson, Zhuoer Lyu et al.

Assessing the exploitability of software vulnerabilities at the time of disclosure is difficult and error-prone, as features extracted via technical analysis by existing metrics are poor predictors for exploit development. Moreover, exploitability assessments suffer from a class bias because "not exploitable" labels could be inaccurate. To overcome these challenges, we propose a new metric, called Expected Exploitability (EE), which reflects, over time, the likelihood that functional exploits will be developed. Key to our solution is a time-varying view of exploitability, a departure from existing metrics. This allows us to learn EE using data-driven techniques from artifacts published after disclosure, such as technical write-ups and proof-of-concept exploits, for which we design novel feature sets. This view also allows us to investigate the effect of the label biases on the classifiers. We characterize the noise-generating process for exploit prediction, showing that our problem is subject to the most challenging type of label noise, and propose techniques to learn EE in the presence of noise. On a dataset of 103,137 vulnerabilities, we show that EE increases precision from 49% to 86% over existing metrics, including two state-of-the-art exploit classifiers, while its precision substantially improves over time. We also highlight the practical utility of EE for predicting imminent exploits and prioritizing critical vulnerabilities. We develop EE into an online platform which is publicly available at https://exploitability.app/.

Yigitcan Kaya, Sanghyun Hong, Tudor Dumitras

Deep learning models often raise privacy concerns as they leak information about their training data. This enables an adversary to determine whether a data point was in a model's training set by conducting a membership inference attack (MIA). Prior work has conjectured that regularization techniques, which combat overfitting, may also mitigate the leakage. While many regularization mechanisms exist, their effectiveness against MIAs has not been studied systematically, and the resulting privacy properties are not well understood. We explore the lower bound for information leakage that practical attacks can achieve. First, we evaluate the effectiveness of 8 mechanisms in mitigating two recent MIAs, on three standard image classification tasks. We find that certain mechanisms, such as label smoothing, may inadvertently help MIAs. Second, we investigate the potential of improving the resilience to MIAs by combining complementary mechanisms. Finally, we quantify the opportunity of future MIAs to compromise privacy by designing a white-box `distance-to-confident' (DtC) metric, based on adversarial sample crafting. Our metric reveals that, even when existing MIAs fail, the training samples may remain distinguishable from test samples. This suggests that regularization mechanisms can provide a false sense of privacy, even when they appear effective against existing MIAs.

Sanghyun Hong, Varun Chandrasekaran, Yiğitcan Kaya et al.

Machine learning algorithms are vulnerable to data poisoning attacks. Prior taxonomies that focus on specific scenarios, e.g., indiscriminate or targeted, have enabled defenses for the corresponding subset of known attacks. Yet, this introduces an inevitable arms race between adversaries and defenders. In this work, we study the feasibility of an attack-agnostic defense relying on artifacts that are common to all poisoning attacks. Specifically, we focus on a common element between all attacks: they modify gradients computed to train the model. We identify two main artifacts of gradients computed in the presence of poison: (1) their $\ell_2$ norms have significantly higher magnitudes than those of clean gradients, and (2) their orientation differs from clean gradients. Based on these observations, we propose the prerequisite for a generic poisoning defense: it must bound gradient magnitudes and minimize differences in orientation. We call this gradient shaping. As an exemplar tool to evaluate the feasibility of gradient shaping, we use differentially private stochastic gradient descent (DP-SGD), which clips and perturbs individual gradients during training to obtain privacy guarantees. We find that DP-SGD, even in configurations that do not result in meaningful privacy guarantees, increases the model's robustness to indiscriminate attacks. It also mitigates worst-case targeted attacks and increases the adversary's cost in multi-poison scenarios. The only attack we find DP-SGD to be ineffective against is a strong, yet unrealistic, indiscriminate attack. Our results suggest that, while we currently lack a generic poisoning defense, gradient shaping is a promising direction for future research.

Sanghyun Hong, Michael Davinroy, Yiğitcan Kaya et al.

New data processing pipelines and novel network architectures increasingly drive the success of deep learning. In consequence, the industry considers top-performing architectures as intellectual property and devotes considerable computational resources to discovering such architectures through neural architecture search (NAS). This provides an incentive for adversaries to steal these novel architectures; when used in the cloud, to provide Machine Learning as a Service, the adversaries also have an opportunity to reconstruct the architectures by exploiting a range of hardware side channels. However, it is challenging to reconstruct novel architectures and pipelines without knowing the computational graph (e.g., the layers, branches or skip connections), the architectural parameters (e.g., the number of filters in a convolutional layer) or the specific pre-processing steps (e.g. embeddings). In this paper, we design an algorithm that reconstructs the key components of a novel deep learning system by exploiting a small amount of information leakage from a cache side-channel attack, Flush+Reload. We use Flush+Reload to infer the trace of computations and the timing for each computation. Our algorithm then generates candidate computational graphs from the trace and eliminates incompatible candidates through a parameter estimation process. We implement our algorithm in PyTorch and Tensorflow. We demonstrate experimentally that we can reconstruct MalConv, a novel data pre-processing pipeline for malware detection, and ProxylessNAS- CPU, a novel network architecture for the ImageNet classification optimized to run on CPUs, without knowing the architecture family. In both cases, we achieve 0% error. These results suggest hardware side channels are a practical attack vector against MLaaS, and more efforts should be devoted to understanding their impact on the security of deep learning systems.

Sanghyun Hong, Pietro Frigo, Yiğitcan Kaya et al.

Deep neural networks (DNNs) have been shown to tolerate "brain damage": cumulative changes to the network's parameters (e.g., pruning, numerical perturbations) typically result in a graceful degradation of classification accuracy. However, the limits of this natural resilience are not well understood in the presence of small adversarial changes to the DNN parameters' underlying memory representation, such as bit-flips that may be induced by hardware fault attacks. We study the effects of bitwise corruptions on 19 DNN models---six architectures on three image classification tasks---and we show that most models have at least one parameter that, after a specific bit-flip in their bitwise representation, causes an accuracy loss of over 90%. We employ simple heuristics to efficiently identify the parameters likely to be vulnerable. We estimate that 40-50% of the parameters in a model might lead to an accuracy drop greater than 10% when individually subjected to such single-bit perturbations. To demonstrate how an adversary could take advantage of this vulnerability, we study the impact of an exemplary hardware fault attack, Rowhammer, on DNNs. Specifically, we show that a Rowhammer enabled attacker co-located in the same physical machine can inflict significant accuracy drops (up to 99%) even with single bit-flip corruptions and no knowledge of the model. Our results expose the limits of DNNs' resilience against parameter perturbations induced by real-world fault attacks. We conclude by discussing possible mitigations and future research directions towards fault attack-resilient DNNs.

Sanghyun Hong, Tae-hoon Kim, Tudor Dumitraş et al.

As there are increasing needs of sharing data for machine learning, there is growing attention for the owners of the data to claim the ownership. Visible watermarking has been an effective way to claim the ownership of visual data, yet the visibly watermarked images are not regarded as a primary source for learning visual recognition models due to the lost visual information by in the watermark and the possibility of an attack to remove the watermarks. To make the watermarked images better suited for machine learning with less risk of removal, we propose DeepStamp, a watermarking framework that, given a watermarking image and a trained network for image classification, learns to synthesize a watermarked image that are human-perceptible, robust to removals, and able to be used as training images for classification with minimal accuracy loss. To achieve the goal, we employ the generative multi-adversarial network (GMAN). In experiments with CIFAR10, we show that the DeepStamp learn to transform a watermark to be embedded in each image and the watermarked images can be used to train networks.

Yigitcan Kaya, Sanghyun Hong, Tudor Dumitras

We characterize a prevalent weakness of deep neural networks (DNNs)---overthinking---which occurs when a DNN can reach correct predictions before its final layer. Overthinking is computationally wasteful, and it can also be destructive when, by the final layer, a correct prediction changes into a misclassification. Understanding overthinking requires studying how each prediction evolves during a DNN's forward pass, which conventionally is opaque. For prediction transparency, we propose the Shallow-Deep Network (SDN), a generic modification to off-the-shelf DNNs that introduces internal classifiers. We apply SDN to four modern architectures, trained on three image classification tasks, to characterize the overthinking problem. We show that SDNs can mitigate the wasteful effect of overthinking with confidence-based early exits, which reduce the average inference cost by more than 50% and preserve the accuracy. We also find that the destructive effect occurs for 50% of misclassifications on natural inputs and that it can be induced, adversarially, with a recent backdooring attack. To mitigate this effect, we propose a new confusion metric to quantify the internal disagreements that will likely lead to misclassifications.

Sanghyun Hong, Michael Davinroy, Yiǧitcan Kaya et al.

Recent work has introduced attacks that extract the architecture information of deep neural networks (DNN), as this knowledge enhances an adversary's capability to conduct black-box attacks against the model. This paper presents the first in-depth security analysis of DNN fingerprinting attacks that exploit cache side-channels. First, we define the threat model for these attacks: our adversary does not need the ability to query the victim model; instead, she runs a co-located process on the host machine victim's deep learning (DL) system is running and passively monitors the accesses of the target functions in the shared framework. Second, we introduce DeepRecon, an attack that reconstructs the architecture of the victim network by using the internal information extracted via Flush+Reload, a cache side-channel technique. Once the attacker observes function invocations that map directly to architecture attributes of the victim network, the attacker can reconstruct the victim's entire network architecture. In our evaluation, we demonstrate that an attacker can accurately reconstruct two complex networks (VGG19 and ResNet50) having observed only one forward propagation. Based on the extracted architecture attributes, we also demonstrate that an attacker can build a meta-model that accurately fingerprints the architecture and family of the pre-trained model in a transfer learning setting. From this meta-model, we evaluate the importance of the observed attributes in the fingerprinting process. Third, we propose and evaluate new framework-level defense techniques that obfuscate our attacker's observations. Our empirical security analysis represents a step toward understanding the DNNs' vulnerability to cache side-channel attacks.

Ali Shafahi, W. Ronny Huang, Mahyar Najibi et al.

Data poisoning is an attack on machine learning models wherein the attacker adds examples to the training set to manipulate the behavior of the model at test time. This paper explores poisoning attacks on neural nets. The proposed attacks use "clean-labels"; they don't require the attacker to have any control over the labeling of training data. They are also targeted; they control the behavior of the classifier on a $\textit{specific}$ test instance without degrading overall classifier performance. For example, an attacker could add a seemingly innocuous image (that is properly labeled) to a training set for a face recognition engine, and control the identity of a chosen person at test time. Because the attacker does not need to control the labeling function, poisons could be entered into the training set simply by leaving them on the web and waiting for them to be scraped by a data collection bot. We present an optimization-based method for crafting poisons, and show that just one single poison image can control classifier behavior when transfer learning is used. For full end-to-end training, we present a "watermarking" strategy that makes poisoning reliable using multiple ($\approx$50) poisoned training instances. We demonstrate our method by generating poisoned frog images from the CIFAR dataset and using them to manipulate image classifiers.

Octavian Suciu, Radu Mărginean, Yiğitcan Kaya et al.

Recent results suggest that attacks against supervised machine learning systems are quite effective, while defenses are easily bypassed by new attacks. However, the specifications for machine learning systems currently lack precise adversary definitions, and the existing attacks make diverse, potentially unrealistic assumptions about the strength of the adversary who launches them. We propose the FAIL attacker model, which describes the adversary's knowledge and control along four dimensions. Our model allows us to consider a wide range of weaker adversaries who have limited control and incomplete knowledge of the features, learning algorithms and training instances utilized. To evaluate the utility of the FAIL model, we consider the problem of conducting targeted poisoning attacks in a realistic setting: the crafted poison samples must have clean labels, must be individually and collectively inconspicuous, and must exhibit a generalized form of transferability, defined by the FAIL model. By taking these constraints into account, we design StingRay, a targeted poisoning attack that is practical against 4 machine learning applications, which use 3 different learning algorithms, and can bypass 2 existing defenses. Conversely, we show that a prior evasion attack is less effective under generalized transferability. Such attack evaluations, under the FAIL adversary model, may also suggest promising directions for future defenses.

Kristián Kozák, Bum Jun Kwon, Doowon Kim et al.

Recent measurements of the Windows code-signing certificate ecosystem have highlighted various forms of abuse that allow malware authors to produce malicious code carrying valid digital signatures. However, the underground trade that allows miscreants to acquire such certificates is not well understood. In this paper, we illuminate two aspects of this trade. First, we investigate 4 leading vendors of Authenticode certificates, we document how they conduct business, and we estimate their market share. Second, we collect a data set of recently signed malware and we use it to study the relationships among malware developers, malware families and the certificates. We also use information from the black market to fingerprint the certificates traded and to identify when the are likely used to sign malware in the wild. Using these methods, we document a shift in the methods that malware authors employ to obtain valid digital signatures. While prior studies have reported the use of code-signing certificates that had been compromised or obtained directly from legitimate Certification Authorities, we observe that, in 2017, these methods have become secondary to purchasing certificates from underground vendors. We also find that the need to bypass platform protections such as Microsoft Defender SmartScreen plays a growing role in driving the demand for Authenticode certificates. Together, these findings suggest that the trade in certificates issued for abuse represents an emerging segment of the underground economy.

Rock Stevens, Octavian Suciu, Andrew Ruef et al.

Governments and businesses increasingly rely on data analytics and machine learning (ML) for improving their competitive edge in areas such as consumer satisfaction, threat intelligence, decision making, and product efficiency. However, by cleverly corrupting a subset of data used as input to a target's ML algorithms, an adversary can perturb outcomes and compromise the effectiveness of ML technology. While prior work in the field of adversarial machine learning has studied the impact of input manipulation on correct ML algorithms, we consider the exploitation of bugs in ML implementations. In this paper, we characterize the attack surface of ML programs, and we show that malicious inputs exploiting implementation bugs enable strictly more powerful attacks than the classic adversarial machine learning techniques. We propose a semi-automated technique, called steered fuzzing, for exploring this attack surface and for discovering exploitable bugs in machine learning programs, in order to demonstrate the magnitude of this threat. As a result of our work, we responsibly disclosed five vulnerabilities, established three new CVE-IDs, and illuminated a common insecure practice across many machine learning systems. Finally, we outline several research directions for further understanding and mitigating this threat.

Bum Jun Kwon, Virinchi Srinivas, Amol Deshpande et al.

The growing commoditization of the underground economy has given rise to malware delivery networks, which charge fees for quickly delivering malware or unwanted software to a large number of hosts. To provide this service, a key method is the orchestration of silent delivery campaigns, which involve a group of downloaders that receive remote commands and that deliver their payloads without any user interaction. These campaigns have not been characterized systematically, unlike other aspects of malware delivery networks. Moreover, silent delivery campaigns can evade detection by relying on inconspicuous downloaders on the client side and on disposable domain names on the server side. We describe Beewolf, a system for detecting silent delivery campaigns from Internet-wide records of download events. The key observation behind our system is that the downloaders involved in these campaigns frequently retrieve payloads in lockstep. Beewolf identifies such locksteps in an unsupervised and deterministic manner. By exploiting novel techniques and empirical observations, Beewolf can operate on streaming data. We utilize Beewolf to study silent delivery campaigns at scale, on a data set of 33.3 million download events. This investigation yields novel findings, e.g. malware distributed through compromised software update channels, a substantial overlap between the delivery ecosystems for malware and unwanted software, and several types of business relationships within these ecosystems. Beewolf achieves over 92% true positives and fewer than 5% false positives. Moreover, Beewolf can detect suspicious downloaders a median of 165 days ahead of existing anti-virus products and payload-hosting domains a median of 196 days ahead of existing blacklists.