As progress in AI continues to advance, it is important to know how advanced systems will make choices and in what ways they may fail. Machines can already outsmart humans in some domains, and understanding how to safely build ones which may have capabilities at or above the human level is of particular concern. One might suspect that artificially generally intelligent (AGI) and artificially superintelligent (ASI) will be systems that humans cannot reliably outsmart. As a challenge to this assumption, this paper presents the Achilles Heel hypothesis which states that even a potentially superintelligent system may nonetheless have stable decision-theoretic delusions which cause them to make irrational decisions in adversarial settings. In a survey of key dilemmas and paradoxes from the decision theory literature, a number of these potential Achilles Heels are discussed in context of this hypothesis. Several novel contributions are made toward understanding the ways in which these weaknesses might be implanted into a system. Read More
#adversarialTag Archives: Adversarial
More than you’ve asked for: A Comprehensive Analysis of Novel Prompt Injection Threats to Application-Integrated Large Language Models
We are currently witnessing dramatic advances in the capabilities of Large Language Models (LLMs). They are already being adopted in practice and integrated into many systems, including integrated development environments (IDEs) and search engines. The functionalities of current LLMs can be modulated via natural language prompts, while their exact internal functionality remains implicit and unassessable. This property, which makes them adaptable to even unseen tasks, might also make them susceptible to targeted adversarial prompting. Recently, several ways to misalign LLMs using Prompt Injection (PI) attacks have been introduced. In such attacks, an adversary can prompt the LLM to produce malicious content or override the original instructions and the employed filtering schemes. Recent work showed that these attacks are hard to mitigate, as state-of-the-art LLMs are instruction-following. So far, these attacks assumed that the adversary is directly prompting the LLM. In this work, we show that augmenting LLMs with retrieval and API calling capabilities (so-called Application-Integrated LLMs) induces a whole new set of attack vectors. These LLMs might process poisoned content retrieved from the Web that contains malicious prompts pre-injected and selected by adversaries. We demonstrate that an attacker can indirectly perform such PI attacks. Based on this key insight, we systematically analyze the resulting threat landscape of Application-Integrated LLMs and discuss a variety of new attack vectors. To demonstrate the practical viability of our attacks, we implemented specific demonstrations of the proposed attacks within synthetic applications. In summary, our work calls for an urgent evaluation of current mitigation techniques and an investigation of whether new techniques are needed to defend LLMs against these threats. Read More
Attacking Machine Learning Systems
The field of machine learning (ML) security—and corresponding adversarial ML—is rapidly advancing as researchers develop sophisticated techniques to perturb, disrupt, or steal the ML model or data. It’s a heady time; because we know so little about the security of these systems, there are many opportunities for new researchers to publish in this field. In many ways, this circumstance reminds me of the cryptanalysis field in the 1990. And there is a lesson in that similarity: the complex mathematical attacks make for good academic papers, but we mustn’t lose sight of the fact that insecure software will be the likely attack vector for most ML systems.
We are amazed by real-world demonstrations of adversarial attacks on ML systems, such as a 3D-printed object that looks like a turtle but is recognized (from any orientation) by the ML system as a gun. Or adding a few stickers that look like smudges to a stop sign so that it is recognized by a state-of-the-art system as a 45 mi/h speed limit sign. But what if, instead, somebody hacked into the system and just switched the labels for “gun” and “turtle” or swapped “stop” and “45 mi/h”? Systems can only match images with human-provided labels, so the software would never notice the switch. That is far easier and will remain a problem even if systems are developed that are robust to those adversarial attacks.
At their core, modern ML systems have complex mathematical models that use training data to become competent at a task. And while there are new risks inherent in the ML model, all of that complexity still runs in software. …. Read More
Machine Generated Text: A Comprehensive Survey of Threat Models and Detection Methods
Advances in natural language generation (NLG) have resulted in machine generated text that is increasingly difficult to distinguish from human authored text. Powerful open-source models are freely available, and user-friendly tools democratizing access to generative models are proliferating. The great potential of state-of-the-art NLG systems is tempered by the multitude of avenues for abuse. Detection of machine generated text is a key countermeasure for reducing abuse of NLG models, with significant technical challenges and numerous open problems. We provide a survey that includes both 1) an extensive analysis of threat models posed by contemporary NLG systems, and 2) the most complete review of machine generated text detection methods to date. This survey places machine generated text within its cybersecurity and social context, and provides strong guidance for future work addressing the most critical threat models, and ensuring detection systems themselves demonstrate trustworthiness through fairness, robustness, and accountability. Read More
ImpNet: Imperceptible and blackbox-undetectable backdoors in compiled neural networks
Early backdoor attacks against machine learning set off an arms race in attack and defence development. Defences have since appeared demonstrating some ability to detect backdoors in models or even remove them. These defences work by inspecting the training data, the model, or the integrity of the training procedure. In this work, we show that backdoors can be added during compilation, circumventing any safeguards in the data preparation and model training stages. As an illustration, the attacker can insert weight-based backdoors during the hardware compilation step that will not be detected by any training or data-preparation process. Next, we demonstrate that some backdoors, such as ImpNet, can only be reliably detected at the stage where they are inserted and removing them anywhere else presents a significant challenge. We conclude that machine-learning model security requires assurance of provenance along the entire technical pipeline, including the data, model architecture, compiler, and hardware specification. Read More
#adversarialWho Are You (I Really Wanna Know)? Detecting Audio DeepFakes Through Vocal Tract Reconstruction
Generative machine learning models have made convincing voice synthesis a reality. While such tools can be extremely useful in applications where people consent to their voices being cloned (e.g., patients losing the ability to speak, actors not wanting to have to redo dialog, etc), they also allow for the creation of nonconsensual content known as deepfakes. This malicious audio is problematic not only because it can convincingly be used to impersonate arbitrary users, but because detecting deepfakes is challenging and generally requires knowledge of the specific deepfake generator. In this paper, we develop a new mechanism for detecting audio deepfakes using techniques from the field of articulatory phonetics. Specifically, we apply fluid dynamics to estimate the arrangement of the human vocal tract during speech generation and show that deepfakes often model impossible or highly-unlikely anatomical arrangements. When parameterized to achieve 99.9% precision, our detection mechanism achieves a recall of 99.5%, correctly identifying all but one deepfake sample in our dataset. We then discuss the limitations of this approach, and how deepfake models fail to reproduce all aspects of speech equally. In so doing, we demonstrate that subtle, but biologically constrained aspects of how humans generate speech are not captured by current models, and can therefore act as a powerful tool to detect audio deepfakes. Read More
You can’t solve AI security problems with more AI
One of the most common proposed solutions to prompt injection attacks (where an AI language model backed system is subverted by a user injecting malicious input—“ignore previous instructions and do this instead”) is to apply more AI to the problem.
I wrote about how I don’t know how to solve prompt injection the other day. I still don’t know how to solve it, but I’m very confident that adding more AI is not the right way to go. Read More
I don’t know how to solve prompt injection
Some extended thoughts about prompt injection attacks against software built on top of AI language models such a GPT-3. This post started as a Twitter thread but I’m promoting it to a full blog entry here.
The more I think about these prompt injection attacks against GPT-3, the more my amusement turns to genuine concern.
I know how to beat XSS, and SQL injection, and so many other exploits.
I have no idea how to reliably beat prompt injection! Read More
Sponge Examples: Energy-Latency Attacks on Neural Networks
The high energy costs of neural network training and inference led to the use of acceleration hardware such as GPUs and TPUs. While such devices enable us to train large-scale neural networks in datacenters and deploy them on edge devices, their designers’ focus so far is on average-case performance. In this work, we introduce a novel threat vector against neural networks whose energy consumption or decision latency are critical. We show how adversaries can exploit carefully-crafted sponge examples, which are inputs designed to maximise energy consumption and latency, to drive machine learning (ML) systems towards their worst-case performance. Sponge examples are, to our knowledge, the first denial-of-service attack against the ML components of such systems. We mount two variants of our sponge attack on a wide range of state-of-the-art neural network models, and find that language models are surprisingly vulnerable. Sponge examples frequently increase both latency and energy consumption of these models by a factor of 30×. Extensive experiments show that our new attack is effective across different hardware platforms (CPU, GPU and an ASIC simulator) on a wide range of different language tasks. On vision tasks, we show that sponge examples can be produced and a latency degradation observed, but the effect is less pronounced. To demonstrate the effectiveness of sponge examples in the real world, we mount an attack against Microsoft Azure’s translator and show an increase of response time from 1ms to 6s (6000×). We conclude by proposing a defense strategy: shifting the analysis of energy consumption in hardware from an average-case to a worst-case perspective. Read More
Manipulating SGC with data ordering attacks
Machine learning is vulnerable to a wide variety of attacks. It is now well understood that by changing the underlying data distribution, an adversary can poison the model trained with it or introduce backdoors. In this paper we present a novel class of training-time attacks that require no changes to the underlying dataset or model architecture, but instead only change the order in which data are supplied to the model. In particular, we find that the attacker can either prevent the model from learning, or poison it to learn behaviours specified by the attacker. Furthermore, we find that even a single adversarially-ordered epoch can be enough to slow down model learning, or even to reset all of the learning progress. Indeed, the attacks presented here are not specific to the model or dataset, but rather target the stochastic nature of modern learning procedures. We extensively evaluate our attacks on computer vision and natural language benchmarks to find that the adversary can disrupt model training and even introduce backdoors. Read More
Rick's Cafe AI 