Papers

Publications


Press [1]

The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks

USENIX Security, 2019.

Nicholas Carlini, Chang Liu, Ulfar Erlingsson, Jernej Kos, Dawn Song

This paper describes a testing methodology for quantitatively assessing the risk that rare or unique training-data sequences are unintentionally memorized by generative sequence models—a common type of machine-learning model. Because such models are sometimes trained on sensitive data (e.g., the text of users' private messages), this methodology can benefit privacy by allowing deep-learning practitioners to select means of training that minimize such memorization.

In experiments, we show that unintended memorization is a persistent, hard-to-avoid issue that can have serious consequences. Specifically, for models trained without consideration of memorization, we describe new, efficient procedures that can extract unique, secret sequences, such as credit card numbers. We show that our testing strategy is a practical and easy-to-use first line of defense, e.g., by describing its application to quantitatively limit data exposure in Google's Smart Compose, a commercial text-completion neural network trained on millions of users' email messages.



Adversarial Examples Are a Natural Consequence of Test Error in Noise

ICML, 2019.

Nic Ford, Justin Gilmer, Nicholas Carlini, Dogus Cubuk

Over the last few years, the phenomenon of adversarial examples --- maliciously constructed inputs that fool trained machine learning models --- has captured the attention of the research community, especially when the adversary is restricted to small modifications of a correctly handled input. Less surprisingly, image classifiers also lack human-level performance on randomly corrupted images, such as images with additive Gaussian noise. In this paper we provide both empirical and theoretical evidence that these are two manifestations of the same underlying phenomenon, establishing close connections between the adversarial robustness and corruption robustness research programs. This suggests that improving adversarial robustness should go hand in hand with improving performance in the presence of more general and realistic image corruptions. Based on our results we recommend that future adversarial defenses consider evaluating the robustness of their methods to distributional shift with benchmarks such as Imagenet-C.



Examples

Imperceptible, Robust, and Targeted Adversarial Examples for Automatic Speech Recognition

ICML, 2019.

Yao Qin, Nicholas Carlini, Ian Goodfellow, Garrison Cottrell, Colin Raffel

Adversarial examples are inputs to machine learning models designed by an adversary to cause an incorrect output. So far, adversarial examples have been studied most extensively in the image domain. In this domain, adversarial examples can be constructed by imperceptibly modifying images to cause misclassification, and are practical in the physical world. In contrast, current targeted adversarial examples applied to speech recognition systems have neither of these properties: humans can easily identify the adversarial perturbations, and they are not effective when played over-the-air. This paper makes advances on both of these fronts. First, we develop effectively imperceptible audio adversarial examples (verified through a human study) by leveraging the psychoacoustic principle of auditory masking, while retaining 100% targeted success rate on arbitrary full-sentence targets. Next, we make progress towards physical-world over-the-air audio adversarial examples by constructing perturbations which remain effective even after applying realistic simulated environmental distortions.



Exploiting Excessive Invariance caused by Norm-Bounded Adversarial Robustness

SafeML ICLR Workshop, 2019.

Jörn-Henrik Jacobsen, Jens Behrmannn, Nicholas Carlini, Florian Tramèr, Nicolas Papernot

Adversarial examples are malicious inputs crafted to cause a model to misclassify them. Their most common instantiation, "perturbation-based" adversarial examples introduce changes to the input that leave its true label unchanged, yet result in a different model prediction. Conversely, "invariance-based" adversarial examples insert changes to the input that leave the model's prediction unaffected despite the underlying input's label having changed.

In this paper, we demonstrate that robustness to perturbation-based adversarial examples is not only insufficient for general robustness, but worse, it can also increase vulnerability of the model to invariance-based adversarial examples. In addition to analytical constructions, we empirically study vision classifiers with state-of-the-art robustness to perturbation-based adversaries constrained by an lp norm. We mount attacks that exploit excessive model invariance in directions relevant to the task, which are able to find adversarial examples within the lp ball. In fact, we find that classifiers trained to be lp-norm robust are more vulnerable to invariance-based adversarial examples than their undefended counterparts.

Excessive invariance is not limited to models trained to be robust to perturbation-based lp-norm adversaries. In fact, we argue that the term adversarial example is used to capture a series of model limitations, some of which may not have been discovered yet. Accordingly, we call for a set of precise definitions that taxonomize and address each of these shortcomings in learning.



Code

On Evaluating Adversarial Robustness

arXiv (unpublished), 2019.

Nicholas Carlini, Anish Athalye, Nicolas Papernot, Wieland Brendel, Jonas Rauber, Dimitris Tsipras, Ian Goodfellow, Aleksander Madry

Correctly evaluating defenses against adversarial examples has proven to be extremely difficult. Despite the significant amount of recent work attempting to design defenses that withstand adaptive attacks, few have succeeded; most papers that propose defenses are quickly shown to be incorrect.

We believe a large contributing factor is the difficulty of performing security evaluations. In this paper, we discuss the methodological foundations, review commonly accepted best practices, and suggest new methods for evaluating defenses to adversarial examples. We hope that both researchers developing defenses as well as readers and reviewers who wish to understand the completeness of an evaluation consider our advice in order to avoid common pitfalls.



Code

Unrestricted Adversarial Examples

arXiv (unpublished), 2018.

Tom B. Brown, Nicholas Carlini, Chiyuan Zhang, Catherine Olsson, Paul Christiano, Ian Goodfellow

We introduce a two-player contest for evaluating the safety and robustness of machine learning systems, with a large prize pool. Unlike most prior work in ML robustness, which studies norm-constrained adversaries, we shift our focus to unconstrained adversaries. Defenders submit machine learning models, and try to achieve high accuracy and coverage on non-adversarial data while making no confident mistakes on adversarial inputs. Attackers try to subvert defenses by finding arbitrary unambiguous inputs where the model assigns an incorrect label with high confidence. We propose a simple unambiguous dataset ("bird-or-bicycle") to use as part of this contest. We hope this contest will help to more comprehensively evaluate the worst-case adversarial risk of machine learning models.



Code, Press [1, 2]

Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples

International Conference on Machine Learning, 2018. Best Paper.

Anish Athalye*, Nicholas Carlini*, and David Wagner

We identify obfuscated gradients, a kind of gradient masking, as a phenomenon that leads to a false sense of security in defenses against adversarial examples. While defenses that cause obfuscated gradients appear to defeat iterative optimization-based attacks, we find defenses relying on this effect can be circumvented. We describe characteristic behaviors of defenses exhibiting the effect, and for each of the three types of obfuscated gradients we discover, we develop attack techniques to overcome it. In a case study, examining non-certified white-box-secure defenses at ICLR 2018, we find obfuscated gradients are a common occurrence, with 7 of 9 defenses relying on obfuscated gradients. Our new attacks successfully circumvent 6 completely, and 1 partially, in the original threat model each paper considers.

* Equal Contribution


Code, Talk, Press [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]

Audio Adversarial Examples: Targeted Attacks on Speech-to-Text

Deep Learning and Security Workshop, 2018. Best Paper.

Nicholas Carlini and David Wagner

We construct targeted audio adversarial examples on automatic speech recognition. Given any audio waveform, we can produce another that is over 99.9% similar, but transcribes as any phrase we choose (recognizing up to 50 characters per second of audio). We apply our white-box iterative optimization-based attack to Mozilla’s implementation DeepSpeech end-to-end, and show it has a 100% success rate. The feasibility of this attack introduce a new domain to study adversarial examples.


Slides, Code

Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection Methods

ACM Workshop on Artificial Intelligence and Security, 2017. Finalist, Best Paper.

Nicholas Carlini and David Wagner

Neural networks are known to be vulnerable to adversarial examples: inputs that are close to natural inputs but classied incorrectly. In order to better understand the space of adversarial examples, we survey ten recent proposals that are designed for detection and compare their efficacy. We show that all can be defeated by constructing new loss functions. We conclude that adversarial examples are significantly harder to detect than previously appreciated, and the properties believed to be intrinsic to adversarial examples are in fact not. Finally, we propose several simple guidelines for evaluating future proposed defenses.


Slides

Adversarial Example Defenses: Ensembles of Weak Defenses are not Strong

USENIX Workshop on Offensive Technologies, 2017.

Warren He, James Wei, Xinyun Chen, Nicholas Carlini, Dawn Song

Ongoing research has proposed several methods to defend neural networks against adversarial examples, many of which researchers have shown to be ineffective. We ask whether a strong defense can be created by combining multiple (possibly weak) defenses. To answer this question, we study three defenses that follow this approach. Two of these are recently proposed defenses that intentionally combine components designed to work well together. A third defense combines three independent defenses. For all the components of these defenses and the combined defenses themselves, we show that an adaptive adversary can create adversarial examples successfully with low distortion. Thus, our work implies that ensemble of weak defenses is not sufficient to provide strong defense against adversarial examples.


Talk, Code

Towards Evaluating the Robustness of Neural Networks

IEEE Symposium on Security and Privacy, 2017. Best Student Paper.

Nicholas Carlini and David Wagner

Neural networks provide state-of-the-art results for most machine learning tasks. Unfortunately, neural networks are vulnerable to adversarial examples: given an input x and any target classification t, it is possible to find a new input x' that is similar to x but classified as t. This makes it difficult to apply neural networks in security-critical areas. Defensive distillation is a recently proposed approach that can take an arbitrary neural network, and increase its robustness, reducing the success rate of current attacks’ ability to find adversarial examples from 95% to 0.5%.

In this paper, we demonstrate that defensive distillation does not significantly increase the robustness of neural networks by introducing three new attack algorithms that are successful on both distilled and undistilled neural networks with 100% probability. Our attacks are tailored to three distance metrics used previously in the literature, and when compared to previous adversarial example generation algorithms, our attacks are often much more effective (and never worse). Furthermore, we propose using high-confidence adversarial examples in a simple transferability test we show can also be used to break defensive distillation. We hope our attacks will be used as a benchmark in future defense attempts to create neural networks that resist adversarial examples.



Talk, Press [1, 2, 3, 4, 5, 6, 7]

Hidden Voice Commands

USENIX Security, 2016. CSAW Best Applied Research Paper.

Nicholas Carlini*, Pratyush Mishra*, Tavish Vaidya*, Yuankai Zhang*, Micah Sherr, Clay Shields, David Wagner, and Wenchao Zhou

Voice interfaces are becoming more ubiquitous and are now the primary input method for many devices. We explore in this paper how they can be attacked with hidden voice commands that are unintelligible to human listeners but which are interpreted as commands by devices.

We evaluate these attacks under two different threat models. In the black-box model, an attacker uses the speech recognition system as an opaque oracle. We show that the adversary can produce difficult to understand commands that are effective against existing systems in the black-box model. Under the white-box model, the attacker has full knowledge of the internals of the speech recognition system and uses it to create attack commands that we demonstrate through user testing are not understandable by humans.

We then evaluate several defenses, including notifying the user when a voice command is accepted; a verbal challenge-response protocol; and a machine learning approach that can detect our attacks with 99.8% accuracy.


* authors listed alphabetically, students appearing first


Talk

Control-Flow Bending: On the Effectiveness of Control-Flow Integrity

USENIX Security, 2015.

Nicholas Carlini, Antonio Barresi, Mathias Payer, Thomas R. Gross and David Wagner

Control-Flow Integrity (CFI) is a defense which prevents control-flow hijacking attacks. While recent research has shown that coarse-grained CFI does not stop attacks, fine-grained CFI is believed to be secure.

We argue that assessing the effectiveness of practical CFI implementations is non-trivial and that common evaluation metrics fail to do so. We then evaluate fully-precise static CFI -- the most restrictive CFI policy that does not break functionality -- and reveal limitations in its security. Using a generalization of non-control-data attacks which we call Control-Flow Bending (CFB), we show how an attacker can leverage a memory corruption vulnerability to achieve Turing-complete computation on memory using just calls to the standard library. We use this attack technique to evaluate fully-precise static CFI on six real binaries and show that in five out of six cases, powerful attacks are still possible. Our results suggest that CFI may not be a reliable defense against memory corruption vulnerabilities.

We further evaluate shadow stacks in combination with CFI and find that their presence for security is necessary: deploying shadow stacks removes arbitrary code execution capabilities of attackers in three of six cases.



Talk

ROP is Still Dangerous: Breaking Modern Defenses

USENIX Security, 2014.

Nicholas Carlini and David Wagner

Return Oriented Programming (ROP) has become the exploitation technique of choice for modern memory-safety vulnerability attacks. Recently, there have been multiple attempts at defenses to prevent ROP attacks. In this paper, we introduce three new attack methods that break many existing ROP defenses. Then we show how to break kBouncer and ROPecker, two recent low-overhead defenses that can be applied to legacy software on existing hardware. We examine several recent ROP attacks seen in the wild and demonstrate that our techniques successfully cloak them so they are not detected by these defenses. Our attacks apply to many CFI-based defenses which we argue are weaker than previously thought. Future defenses will need to take our attacks into account.


Talk

Improved Support for Machine-Assisted Ballot-Level Audits

USENIX Journal of Election Technology and Systems (JETS), Volume 1 Issue 1. Presented at EVT/WOTE 2013.

Eric Kim, Nicholas Carlini, Andrew Chang, George Yiu, Kai Wang, and David Wagner

This paper studies how to provide support for ballot-level post-election audits. Informed by our work supporting pilots of these audits in several California counties, we identify gaps in current technology in tools for this task: we need better ways to count voted ballots (from scanned images) without access to scans of blank, unmarked ballots; and we need improvements to existing techniques that help them scale better to large, complex elections. We show how to meet these needs and use our system to successfully process ballots from 11 California counties, in support of the pilot audit program. Our new techniques yield order-of-magnitude speedups compared to the previous system, and enable us to successfully process some elections that would not have reasonably feasible without these techniques.


Talk

Operator-Assisted Tabulation of Optical Scan Ballots

EVT/WOTE, 2012.

Kai Wang, Eric Kim, Nicholas Carlini, Ivan Motyashov, Daniel Nguyen, and David Wagner

We present OpenCount: a system that tabulates scanned ballots from an election by combining computer vision algorithms with focused operator assistance. OpenCount is designed to support risk-limiting audits and to be scalable to large elections, robust to conditions encountered using typical scanner hardware, and general to a wide class of ballot types--all without the need for integration with any vendor systems. To achieve these goals, we introduce a novel operator-in-the-loop computer vision pipeline for automatically processing scanned ballots while allowing the operator to intervene in a simple, intuitive manner. We evaluate our system on data collected from five risk-limiting audit pilots conducted in California in 2011.


Talk

An Evaluation of the Google Chrome Extension Security Architecture

USENIX Security, 2012.

Nicholas Carlini, Adrienne Porter Felt, and David Wagner

Vulnerabilities in browser extensions put users at risk by providing a way for website and network attackers to gain access to users’ private data and credentials. Extensions can also introduce vulnerabilities into the websites that they modify. In 2009, Google Chrome introduced a new extension platform with several features intended to prevent and mitigate extension vulnerabilities: strong isolation between websites and extensions, privilege separation within an extension, and an extension permission system. We performed a security review of 100 Chrome extensions and found 70 vulnerabilities across 40 extensions. Given these vulnerabilities, we evaluate how well each of the security mechanisms defends against extension vulnerabilities. We find that the mechanisms mostly succeed at preventing direct web attacks on extensions, but new security mechanisms are needed to protect users from network attacks on extensions, website metadata attacks on extensions, and vulnerabilities that extensions add to websites. We propose and evaluate additional defenses, and we conclude that banning HTTP scripts and inline scripts would prevent 47 of the 50 most severe vulnerabilities with only modest impact on developers.


Short Papers


A critique of the DeepSec Platform for Security Analysis of Deep Learning Models

arXiv short paper, 2019.

Nicholas Carlini

At IEEE S&P 2019, the paper "DeepSec: A Uniform Platform for Security Analysis of Deep Learning Model" aims to to "systematically evaluate the existing adversarial attack and defense methods." While the paper's goals are laudable, it fails to achieve them and presents results that are fundamentally flawed and misleading. We explain the flaws in the DeepSec work, along with how its analysis fails to meaningfully evaluate the various attacks and defenses. Specifically, DeepSec (1) evaluates each defense obliviously, using attacks crafted against undefended models; (2) evaluates attacks and defenses using incorrect implementations that greatly under-estimate their effectiveness; (3) evaluates the robustness of each defense as an average, not based on the most effective attack against that defense; (4) performs several statistical analyses incorrectly and fails to report variance; and, (5) as a result of these errors draws invalid conclusions and makes sweeping generalizations.



On the Robustness of the CVPR 2018 White-Box Adversarial Example Defenses

Computer Vision: Challenges and Opportunities for Privacy and Security, 2018.

Anish Athalye and Nicholas Carlini

Neural networks are known to be vulnerable to adversarial examples. In this note, we evaluate the two white-box defenses that appeared at CVPR 2018 and find they are ineffective: when applying existing techniques, we can reduce the accuracy of the defended models to 0%.


MagNet and "Efficient Defenses Against Adversarial Attacks" are Not Robust to Adversarial Examples

arXiv short paper, 2017.

Nicholas Carlini and David Wagner

MagNet and "Efficient Defenses..." were recently proposed as a defense to adversarial examples. We find that we can construct adversarial examples that defeat these defenses with only a slight increase in distortion.


Defensive Distillation is Not Robust to Adoversarial Examples

arXiv short paper, 2016.

Nicholas Carlini and David Wagner

We show that defensive distillation is not secure: it is no more resistant to targeted misclassification attacks than unprotected neural networks.