Dawn Song

Clark Barrett, Brad Boyd, Elie Burzstein et al. · berkeley

Every major technical invention resurfaces the dual-use dilemma -- the new technology has the potential to be used for good as well as for harm. Generative AI (GenAI) techniques, such as large language models (LLMs) and diffusion models, have shown remarkable capabilities (e.g., in-context learning, code-completion, and text-to-image generation and editing). However, GenAI can be used just as well by attackers to generate new attacks and increase the velocity and efficacy of existing attacks. This paper reports the findings of a workshop held at Google (co-organized by Stanford University and the University of Wisconsin-Madison) on the dual-use dilemma posed by GenAI. This paper is not meant to be comprehensive, but is rather an attempt to synthesize some of the interesting findings from the workshop. We discuss short-term and long-term goals for the community on this topic. We hope this paper provides both a launching point for a discussion on this important topic as well as interesting problems that the research community can work to address.

Rishub Tamirisa, Bhrugu Bharathi, Long Phan et al. · cmu

Rapid advances in the capabilities of large language models (LLMs) have raised widespread concerns regarding their potential for malicious use. Open-weight LLMs present unique challenges, as existing safeguards lack robustness to tampering attacks that modify model weights. For example, recent works have demonstrated that refusal and unlearning safeguards can be trivially removed with a few steps of fine-tuning. These vulnerabilities necessitate new approaches for enabling the safe release of open-weight LLMs. We develop a method, called TAR, for building tamper-resistant safeguards into open-weight LLMs such that adversaries cannot remove the safeguards even after hundreds of steps of fine-tuning. In extensive evaluations and red teaming analyses, we find that our method greatly improves tamper-resistance while preserving benign capabilities. Our results demonstrate that progress on tamper-resistance is possible, opening up a promising new avenue to improve the safety and security of open-weight LLMs.

Ishir Garg, Neel Kolhe, Xuandong Zhao et al. · berkeley

Large language models (LLMs) have demonstrated significant advancements in reasoning and code generation. However, efficiently creating new benchmarks to evaluate these capabilities remains a challenge. Traditional benchmark creation relies on manual human effort, a process that is both expensive and time-consuming. Furthermore, existing benchmarks often contaminate LLM training data, necessitating novel and diverse benchmarks to accurately assess their genuine capabilities. This work introduces InfoSynth, a novel framework for automatically generating and evaluating reasoning benchmarks guided by information-theoretic principles. We propose metrics based on KL-divergence and entropy to quantify benchmark novelty and diversity without relying on costly model evaluations. Building on this framework, we develop an end-to-end pipeline that synthesizes robust Python coding problems from seed datasets using genetic algorithms and iterative code feedback. Our method generates accurate test cases and solutions to new problems 97% of the time, and the synthesized benchmarks consistently exhibit higher novelty and diversity compared to their seed datasets. Moreover, our algorithm provides a method for controlling the novelty/diversity and difficulty of generated problems. InfoSynth offers a scalable, self-verifying pipeline for constructing high-quality, novel and diverse benchmarks for LLMs. Project Page: https://ishirgarg.github.io/infosynth_web/

Zhaorun Chen, Zhen Xiang, Chaowei Xiao et al.

LLM agents have demonstrated remarkable performance across various applications, primarily due to their advanced capabilities in reasoning, utilizing external knowledge and tools, calling APIs, and executing actions to interact with environments. Current agents typically utilize a memory module or a retrieval-augmented generation (RAG) mechanism, retrieving past knowledge and instances with similar embeddings from knowledge bases to inform task planning and execution. However, the reliance on unverified knowledge bases raises significant concerns about their safety and trustworthiness. To uncover such vulnerabilities, we propose a novel red teaming approach AgentPoison, the first backdoor attack targeting generic and RAG-based LLM agents by poisoning their long-term memory or RAG knowledge base. In particular, we form the trigger generation process as a constrained optimization to optimize backdoor triggers by mapping the triggered instances to a unique embedding space, so as to ensure that whenever a user instruction contains the optimized backdoor trigger, the malicious demonstrations are retrieved from the poisoned memory or knowledge base with high probability. In the meantime, benign instructions without the trigger will still maintain normal performance. Unlike conventional backdoor attacks, AgentPoison requires no additional model training or fine-tuning, and the optimized backdoor trigger exhibits superior transferability, in-context coherence, and stealthiness. Extensive experiments demonstrate AgentPoison's effectiveness in attacking three types of real-world LLM agents: RAG-based autonomous driving agent, knowledge-intensive QA agent, and healthcare EHRAgent. On each agent, AgentPoison achieves an average attack success rate higher than 80% with minimal impact on benign performance (less than 1%) with a poison rate less than 0.1%.

Feiyang Kang, Yifan Sun, Bingbing Wen et al.

Domain reweighting is an emerging research area aimed at adjusting the relative weights of different data sources to improve the effectiveness and efficiency of LLM pre-training. We show that data mixtures that perform well at smaller scales may not retain their advantage at larger scales, challenging the existing practice of determining competitive mixtures in small-scale experiments and directly applying them at much larger scales. To address this, we propose AutoScale, a two-stage, scale-aware data composition framework. First, AutoScale fits a parametric model that predicts the model's loss under different data compositions, then uses it to find an approximate best allocation at smaller, more manageable budgets. Next, leveraging a novel theoretical analysis of how optimal compositions evolve with scale, AutoScale extrapolates that composition to larger budgets without further retraining. Empirically, AutoScale accelerates convergence and improves downstream performance. For instance, when pre-training GPT-2 Large, it achieves a 28% faster perplexity reduction than baselines and up to a 38% speed-up over unweighted training, while yielding best-average results on various downstream tasks. Overall, our findings illustrate how domain importance shifts with training scale, underscoring the need for scale-dependent data curation in LLM training. Our code is open-sourced.

Xuandong Zhao, Will Cai, Tianneng Shi et al. · berkeley

Existing training-time safety alignment techniques for large language models (LLMs) remain vulnerable to jailbreak attacks. Direct preference optimization (DPO), a widely deployed alignment method, exhibits limitations in both experimental and theoretical contexts as its loss function proves suboptimal for refusal learning. Through gradient-based analysis, we identify these shortcomings and propose an improved safety alignment that disentangles DPO objectives into two components: (1) robust refusal training, which encourages refusal even when partial unsafe generations are produced, and (2) targeted unlearning of harmful knowledge. This approach significantly increases LLM robustness against a wide range of jailbreak attacks, including prefilling, suffix, and multi-turn attacks across both in-distribution and out-of-distribution scenarios. Furthermore, we introduce a method to emphasize critical refusal tokens by incorporating a reward-based token-level weighting mechanism for refusal learning, which further improves the robustness against adversarial exploits. Our research also suggests that robustness to jailbreak attacks is correlated with token distribution shifts in the training process and internal representations of refusal and harmful tokens, offering valuable directions for future research in LLM safety alignment. The code is available at https://github.com/wicai24/DOOR-Alignment

Myeongseob Ko, Henry Li, Zhun Wang et al.

Large-scale generative models have shown impressive image-generation capabilities, propelled by massive data. However, this often inadvertently leads to the generation of harmful or inappropriate content and raises copyright concerns. Driven by these concerns, machine unlearning has become crucial to effectively purge undesirable knowledge from models. While existing literature has studied various unlearning techniques, these often suffer from either poor unlearning quality or degradation in text-image alignment after unlearning, due to the competitive nature of these objectives. To address these challenges, we propose a framework that seeks an optimal model update at each unlearning iteration, ensuring monotonic improvement on both objectives. We further derive the characterization of such an update. In addition, we design procedures to strategically diversify the unlearning and remaining datasets to boost performance improvement. Our evaluation demonstrates that our method effectively removes target classes from recent diffusion-based generative models and concepts from stable diffusion models while maintaining close alignment with the models' original trained states, thus outperforming state-of-the-art baselines. Our code will be made available at https://github.com/reds-lab/Restricted_gradient_diversity_unlearning.git.

Tanusree Sharma, Yujin Potter, Zachary Kilhoffer et al.

How AI models should deal with political topics has been discussed, but it remains challenging and requires better governance. This paper examines the governance of large language models through individual and collective deliberation, focusing on politically sensitive videos. We conducted a two-step study: interviews with 10 journalists established a baseline understanding of expert video interpretation; 114 individuals through deliberation using InclusiveAI, a platform that facilitates democratic decision-making through decentralized autonomous organization (DAO) mechanisms. Our findings reveal distinct differences in interpretative priorities: while experts emphasized emotion and narrative, the general public prioritized factual clarity, objectivity, and emotional neutrality. Furthermore, we examined how different governance mechanisms - quadratic vs. weighted voting and equal vs. 20/80 voting power - shape users' decision-making regarding AI behavior. Results indicate that voting methods significantly influence outcomes, with quadratic voting reinforcing perceptions of liberal democracy and political equality. Our study underscores the necessity of selecting appropriate governance mechanisms to better capture user perspectives and suggests decentralized AI governance as a potential way to facilitate broader public engagement in AI development, ensuring that varied perspectives meaningfully inform design decisions.

Zhe Ye, Zhengxu Yan, Jingxuan He et al.

Large language models (LLMs) are increasingly integrated in software development, but ensuring correctness in LLM-generated code remains challenging and often requires costly manual review. Verifiable code generation -- jointly generating code, specifications, and proofs of code-specification alignment -- offers a promising path to address this limitation and further unleash LLMs' benefits in coding. Yet, there exists a significant gap in evaluation: current benchmarks often focus on only individual components rather than providing a holistic evaluation framework of all tasks. In this paper, we introduce Verina (Verifiable Code Generation Arena), a high-quality benchmark enabling a comprehensive and modular evaluation of code, specification, and proof generation as well as their compositions. Verina consists of 189 manually curated coding tasks in Lean, with detailed problem descriptions, reference implementations, formal specifications, and extensive test suites. Our extensive evaluation of state-of-the-art LLMs reveals significant challenges in verifiable code generation, especially in proof generation, underscoring the need for improving LLM-based theorem provers in verification domains. The best model, OpenAI o4-mini, achieves a 61.4\% code correctness rate, 51.0\% for specification soundness and completeness, and a mere 3.6\% proof success rate (based on one trial per task). We hope Verina will catalyze progress in verifiable code generation by providing a rigorous and comprehensive benchmark. We release our dataset on https://huggingface.co/datasets/sunblaze-ucb/verina and our evaluation code on https://github.com/sunblaze-ucb/verina.

Haoyue Bai, Yiyou Sun, Wenjie Hu et al.

Large Language Models (LLMs) display strikingly different generalization behaviors: supervised fine-tuning (SFT) often narrows capability, whereas reinforcement-learning (RL) tuning tends to preserve it. The reasons behind this divergence remain unclear, as prior studies have largely relied on coarse accuracy metrics. We address this gap by introducing a novel benchmark that decomposes reasoning into atomic core skills such as calculation, fact retrieval, simulation, enumeration, and diagnostic, providing a concrete framework for addressing the fundamental question of what constitutes reasoning in LLMs. By isolating and measuring these core skills, the benchmark offers a more granular view of how specific cognitive abilities emerge, transfer, and sometimes collapse during post-training. Combined with analyses of low-level statistical patterns such as distributional divergence and parameter statistics, it enables a fine-grained study of how generalization evolves under SFT and RL across mathematical, scientific reasoning, and non-reasoning tasks. Our meta-probing framework tracks model behavior at different training stages and reveals that RL-tuned models maintain more stable behavioral profiles and resist collapse in reasoning skills, whereas SFT models exhibit sharper drift and overfit to surface patterns. This work provides new insights into the nature of reasoning in LLMs and points toward principles for designing training strategies that foster broad, robust generalization.

Shiyu Tang, Ruihao Gong, Yan Wang et al.

Deep neural networks (DNNs) are vulnerable to adversarial noises, which motivates the benchmark of model robustness. Existing benchmarks mainly focus on evaluating defenses, but there are no comprehensive studies of how architecture design and training techniques affect robustness. Comprehensively benchmarking their relationships is beneficial for better understanding and developing robust DNNs. Thus, we propose RobustART, the first comprehensive Robustness investigation benchmark on ImageNet regarding ARchitecture design (49 human-designed off-the-shelf architectures and 1200+ networks from neural architecture search) and Training techniques (10+ techniques, e.g., data augmentation) towards diverse noises (adversarial, natural, and system noises). Extensive experiments substantiated several insights for the first time, e.g., (1) adversarial training is effective for the robustness against all noises types for Transformers and MLP-Mixers; (2) given comparable model sizes and aligned training settings, CNNs > Transformers > MLP-Mixers on robustness against natural and system noises; Transformers > MLP-Mixers > CNNs on adversarial robustness; (3) for some light-weight architectures, increasing model sizes or using extra data cannot improve robustness. Our benchmark presents: (1) an open-source platform for comprehensive robustness evaluation; (2) a variety of pre-trained models to facilitate robustness evaluation; and (3) a new view to better understand the mechanism towards designing robust DNNs. We will continuously develop to this ecosystem for the community.

Peng Gao, Xiaoyuan Liu, Edward Choi et al.

To remain aware of the fast-evolving cyber threat landscape, open-source Cyber Threat Intelligence (OSCTI) has received growing attention from the community. Commonly, knowledge about threats is presented in a vast number of OSCTI reports. Despite the pressing need for high-quality OSCTI, existing OSCTI gathering and management platforms, however, have primarily focused on isolated, low-level Indicators of Compromise. On the other hand, higher-level concepts (e.g., adversary tactics, techniques, and procedures) and their relationships have been overlooked, which contain essential knowledge about threat behaviors that is critical to uncovering the complete threat scenario. To bridge the gap, we propose SecurityKG, a system for automated OSCTI gathering and management. SecurityKG collects OSCTI reports from various sources, uses a combination of AI and NLP techniques to extract high-fidelity knowledge about threat behaviors, and constructs a security knowledge graph. SecurityKG also provides a UI that supports various types of interactivity to facilitate knowledge graph exploration.

Peng Gao, Fei Shao, Xiaoyuan Liu et al.

Log-based cyber threat hunting has emerged as an important solution to counter sophisticated cyber attacks. However, existing approaches require non-trivial efforts of manual query construction and have overlooked the rich external knowledge about threat behaviors provided by open-source Cyber Threat Intelligence (OSCTI). To bridge the gap, we build ThreatRaptor, a system that facilitates cyber threat hunting in computer systems using OSCTI. Built upon mature system auditing frameworks, ThreatRaptor provides (1) an unsupervised, light-weight, and accurate NLP pipeline that extracts structured threat behaviors from unstructured OSCTI text, (2) a concise and expressive domain-specific query language, TBQL, to hunt for malicious system activities, (3) a query synthesis mechanism that automatically synthesizes a TBQL query from the extracted threat behaviors, and (4) an efficient query execution engine to search the big system audit logging data.

Peng Gao, Fei Shao, Xiaoyuan Liu et al.

Log-based cyber threat hunting has emerged as an important solution to counter sophisticated attacks. However, existing approaches require non-trivial efforts of manual query construction and have overlooked the rich external threat knowledge provided by open-source Cyber Threat Intelligence (OSCTI). To bridge the gap, we propose ThreatRaptor, a system that facilitates threat hunting in computer systems using OSCTI. Built upon system auditing frameworks, ThreatRaptor provides (1) an unsupervised, light-weight, and accurate NLP pipeline that extracts structured threat behaviors from unstructured OSCTI text, (2) a concise and expressive domain-specific query language, TBQL, to hunt for malicious system activities, (3) a query synthesis mechanism that automatically synthesizes a TBQL query for hunting, and (4) an efficient query execution engine to search the big audit logging data. Evaluations on a broad set of attack cases demonstrate the accuracy and efficiency of ThreatRaptor in practical threat hunting.

Sayash Kapoor, Rishi Bommasani, Kevin Klyman et al.

Foundation models are powerful technologies: how they are released publicly directly shapes their societal impact. In this position paper, we focus on open foundation models, defined here as those with broadly available model weights (e.g. Llama 2, Stable Diffusion XL). We identify five distinctive properties (e.g. greater customizability, poor monitoring) of open foundation models that lead to both their benefits and risks. Open foundation models present significant benefits, with some caveats, that span innovation, competition, the distribution of decision-making power, and transparency. To understand their risks of misuse, we design a risk assessment framework for analyzing their marginal risk. Across several misuse vectors (e.g. cyberattacks, bioweapons), we find that current research is insufficient to effectively characterize the marginal risk of open foundation models relative to pre-existing technologies. The framework helps explain why the marginal risk is low in some cases, clarifies disagreements about misuse risks by revealing that past work has focused on different subsets of the framework with different assumptions, and articulates a way forward for more constructive debate. Overall, our work helps support a more grounded assessment of the societal impact of open foundation models by outlining what research is needed to empirically validate their theoretical benefits and risks.

Yoshua Bengio, Sören Mindermann, Daniel Privitera et al. · eth-zurich, mit

The first International AI Safety Report comprehensively synthesizes the current evidence on the capabilities, risks, and safety of advanced AI systems. The report was mandated by the nations attending the AI Safety Summit in Bletchley, UK. Thirty nations, the UN, the OECD, and the EU each nominated a representative to the report's Expert Advisory Panel. A total of 100 AI experts contributed, representing diverse perspectives and disciplines. Led by the report's Chair, these independent experts collectively had full discretion over the report's content.

Yupei Liu, Yuqi Jia, Jinyuan Jia et al.

LLM-integrated applications and agents are vulnerable to prompt injection attacks, where an attacker injects prompts into their inputs to induce attacker-desired outputs. A detection method aims to determine whether a given input is contaminated by an injected prompt. However, existing detection methods have limited effectiveness against state-of-the-art attacks, let alone adaptive ones. In this work, we propose DataSentinel, a game-theoretic method to detect prompt injection attacks. Specifically, DataSentinel fine-tunes an LLM to detect inputs contaminated with injected prompts that are strategically adapted to evade detection. We formulate this as a minimax optimization problem, with the objective of fine-tuning the LLM to detect strong adaptive attacks. Furthermore, we propose a gradient-based method to solve the minimax optimization problem by alternating between the inner max and outer min problems. Our evaluation results on multiple benchmark datasets and LLMs show that DataSentinel effectively detects both existing and adaptive prompt injection attacks.

Jiawei Zhang, Chejian Xu, Yu Gai et al.

This paper introduces KnowHalu, a novel approach for detecting hallucinations in text generated by large language models (LLMs), utilizing step-wise reasoning, multi-formulation query, multi-form knowledge for factual checking, and fusion-based detection mechanism. As LLMs are increasingly applied across various domains, ensuring that their outputs are not hallucinated is critical. Recognizing the limitations of existing approaches that either rely on the self-consistency check of LLMs or perform post-hoc fact-checking without considering the complexity of queries or the form of knowledge, KnowHalu proposes a two-phase process for hallucination detection. In the first phase, it identifies non-fabrication hallucinations--responses that, while factually correct, are irrelevant or non-specific to the query. The second phase, multi-form based factual checking, contains five key steps: reasoning and query decomposition, knowledge retrieval, knowledge optimization, judgment generation, and judgment aggregation. Our extensive evaluations demonstrate that KnowHalu significantly outperforms SOTA baselines in detecting hallucinations across diverse tasks, e.g., improving by 15.65% in QA tasks and 5.50% in summarization tasks, highlighting its effectiveness and versatility in detecting hallucinations in LLM-generated content.

Jiachen T. Wang, Dawn Song, James Zou et al.

Traditional data influence estimation methods, like influence function, assume that learning algorithms are permutation-invariant with respect to training data. However, modern training paradigms, especially for foundation models using stochastic algorithms and multi-stage curricula, are sensitive to data ordering, thus violating this assumption. This mismatch renders influence functions inadequate for answering a critical question in machine learning: How can we capture the dependence of data influence on the optimization trajectory during training? To address this gap, we formalize the concept of trajectory-specific leave-one-out (LOO) influence, which quantifies the impact of removing a data point from a specific iteration during training, accounting for the exact sequence of data encountered and the model's optimization trajectory. However, exactly evaluating the trajectory-specific LOO presents a significant computational challenge. To address this, we propose data value embedding, a novel technique enabling efficient approximation of trajectory-specific LOO. Specifically, we compute a training data embedding that encapsulates the cumulative interactions between data and the evolving model parameters. The LOO can then be efficiently approximated through a simple dot-product between the data value embedding and the gradient of the given test data. As data value embedding captures training data ordering, it offers valuable insights into model training dynamics. In particular, we uncover distinct phases of data influence, revealing that data points in the early and late stages of training exert a greater impact on the final model. These insights translate into actionable strategies for managing the computational overhead of data selection by strategically timing the selection process, potentially opening new avenues in data curation research.

Yujin Potter, Wenbo Guo, Zhun Wang et al.

The impact of frontier AI in cybersecurity is rapidly increasing. In this paper, we comprehensively analyze this trend through three distinct lenses: a quantitative benchmark analysis, a literature review, and an expert survey. We find that while AI is already widely used in attacks, its application in defense remains limited, especially in remediation and deployment. Aligned with these analyses, experts expect AI to continue favoring attackers over defenders, though the gap will gradually narrow. These findings underscore the urgent need to mitigate frontier AI's risks while closely monitoring emerging capabilities. We provide concrete calls-to-action regarding: the construction of new cybersecurity benchmarks, the development of AI agents for defense, the design of provably secure AI agents, the improvement of pre-deployment security testing and transparency, and the strengthening of user-oriented education and defenses. Our paper summary and blog are available at https://rdi.berkeley.edu/frontier-ai-impact-on-cybersecurity/.

Chenguang Wang, Ruoxi Jia, Xin Liu et al.

Pre-training image representations from the raw text about images enables zero-shot vision transfer to downstream tasks. Through pre-training on millions of samples collected from the internet, multimodal foundation models, such as CLIP, produce state-of-the-art zero-shot results that often reach competitiveness with fully supervised methods without the need for task-specific training. Besides the encouraging performance on classification accuracy, it is reported that these models close the robustness gap by matching the performance of supervised models trained on ImageNet under natural distribution shift. Because robustness is critical to real-world applications, especially safety-critical ones, in this paper, we present a comprehensive evaluation based on a large-scale robustness benchmark covering 7 natural, 3 synthetic distribution shifts, and 11 adversarial attacks. We use CLIP as a pilot study. We show that CLIP leads to a significant robustness drop compared to supervised ImageNet models on our benchmark, especially under synthetic distribution shift and adversarial attacks. Furthermore, data overlap analysis suggests that the observed robustness under natural distribution shifts could be attributed, at least in part, to data overlap. In summary, our evaluation shows a comprehensive evaluation of robustness is necessary; and there is a significant need to improve the robustness of zero-shot multimodal models.

Daniel Thilo Schroeder, Meeyoung Cha, Andrea Baronchelli et al.

Public opinion manipulation has entered a new phase, amplifying its roots in rhetoric and propaganda. Advances in large language models (LLMs) and autonomous agents now let influence campaigns reach unprecedented scale and precision. Researchers warn AI could foster mass manipulation. Generative tools can expand propaganda output without sacrificing credibility and inexpensively create election falsehoods that are rated as more human-like than those written by humans. Techniques meant to refine AI reasoning, such as chain-of-thought prompting, can just as effectively be used to generate more convincing falsehoods. Enabled by these capabilities, another disruptive threat is emerging: swarms of collaborative, malicious AI agents. Fusing LLM reasoning with multi-agent architectures, these systems are capable of coordinating autonomously, infiltrating communities, and fabricating consensus cheaply. By adaptively mimicking human social dynamics, they threaten democracy.

Zhun Wang, Tianneng Shi, Jingxuan He et al.

AI agents have significant potential to reshape cybersecurity, making a thorough assessment of their capabilities critical. However, existing evaluations fall short, because they are based on small-scale benchmarks and only measure static outcomes, failing to capture the full, dynamic range of real-world security challenges. To address these limitations, we introduce CyberGym, a large-scale benchmark featuring 1,507 real-world vulnerabilities across 188 software projects. Adjustable to different vulnerability analysis settings, CyberGym primarily tasks agents with generating a proof-of-concept test that reproduces a vulnerability, given only its text description and the corresponding codebase. Our extensive evaluation highlights that CyberGym effectively differentiates agents' and models' cybersecurity capabilities. Even the top-performing combinations only achieve a ~20% success rate, demonstrating the overall difficulty of CyberGym. Beyond static benchmarking, we show that CyberGym leads to the discovery of 35 zero-day vulnerabilities and 17 historically incomplete patches. These results underscore that CyberGym is not only a robust benchmark for measuring AI's progress in cybersecurity but also a platform for creating direct, real-world security impact.

Bochuan Cao, Jinyuan Jia, Chuxuan Hu et al.

Backdoor attacks aim to inject a backdoor into a classifier such that it predicts any input with an attacker-chosen backdoor trigger as an attacker-chosen target class. Existing backdoor attacks require either retraining the classifier with some clean data or modifying the model's architecture. As a result, they are 1) not applicable when clean data is unavailable, 2) less efficient when the model is large, and 3) less stealthy due to architecture changes. In this work, we propose DFBA, a novel retraining-free and data-free backdoor attack without changing the model architecture. Technically, our proposed method modifies a few parameters of a classifier to inject a backdoor. Through theoretical analysis, we verify that our injected backdoor is provably undetectable and unremovable by various state-of-the-art defenses under mild assumptions. Our evaluation on multiple datasets further demonstrates that our injected backdoor: 1) incurs negligible classification loss, 2) achieves 100% attack success rates, and 3) bypasses six existing state-of-the-art defenses. Moreover, our comparison with a state-of-the-art non-data-free backdoor attack shows our attack is more stealthy and effective against various defenses while achieving less classification accuracy loss.

Jiachen T. Wang, Tong Wu, Kaifeng Lyu et al.

Data teams at frontier AI companies routinely train small proxy models to make critical decisions about pretraining data recipes for full-scale training runs. However, the community has a limited understanding of whether and when conclusions drawn from small-scale experiments reliably transfer to full-scale model training. In this work, we uncover a subtle yet critical issue in the standard experimental protocol for data recipe assessment: the use of identical small-scale model training configurations across all data recipes in the name of "fair" comparison. We show that the experiment conclusions about data quality can flip with even minor adjustments to training hyperparameters, as the optimal training configuration is inherently data-dependent. Moreover, this fixed-configuration protocol diverges from full-scale model development pipelines, where hyperparameter optimization is a standard step. Consequently, we posit that the objective of data recipe assessment should be to identify the recipe that yields the best performance under data-specific tuning. To mitigate the high cost of hyperparameter tuning, we introduce a simple patch to the evaluation protocol: using reduced learning rates for proxy model training. We show that this approach yields relative performance that strongly correlates with that of fully tuned large-scale LLM pretraining runs. Theoretically, we prove that for random-feature models, this approach preserves the ordering of datasets according to their optimal achievable loss. Empirically, we validate this approach across 23 data recipes covering four critical dimensions of data curation, demonstrating dramatic improvements in the reliability of small-scale experiments.

Jiachen T. Wang, Prateek Mittal, Dawn Song et al.

Data Shapley provides a principled framework for attributing data's contribution within machine learning contexts. However, existing approaches require re-training models on different data subsets, which is computationally intensive, foreclosing their application to large-scale models. Furthermore, they produce the same attribution score for any models produced by running the learning algorithm, meaning they cannot perform targeted attribution towards a specific model obtained from a single run of the algorithm. This paper introduces In-Run Data Shapley, which addresses these limitations by offering scalable data attribution for a target model of interest. In its most efficient implementation, our technique incurs negligible additional runtime compared to standard model training. This dramatic efficiency improvement makes it possible to perform data attribution for the foundation model pretraining stage for the first time. We present several case studies that offer fresh insights into pretraining data's contribution and discuss their implications for copyright in generative AI and pretraining data curation.

Zhen Xiang, Linzhi Zheng, Yanjie Li et al.

The rapid advancement of large language model (LLM) agents has raised new concerns regarding their safety and security. In this paper, we propose GuardAgent, the first guardrail agent to protect target agents by dynamically checking whether their actions satisfy given safety guard requests. Specifically, GuardAgent first analyzes the safety guard requests to generate a task plan, and then maps this plan into guardrail code for execution. By performing the code execution, GuardAgent can deterministically follow the safety guard request and safeguard target agents. In both steps, an LLM is utilized as the reasoning component, supplemented by in-context demonstrations retrieved from a memory module storing experiences from previous tasks. In addition, we propose two novel benchmarks: EICU-AC benchmark to assess the access control for healthcare agents and Mind2Web-SC benchmark to evaluate the safety policies for web agents. We show that GuardAgent effectively moderates the violation actions for different types of agents on these two benchmarks with over 98% and 83% guardrail accuracies, respectively. Project page: https://guardagent.github.io/

Dan Hendrycks, Andy Zou, Mantas Mazeika et al.

In real-world applications of machine learning, reliable and safe systems must consider measures of performance beyond standard test set accuracy. These other goals include out-of-distribution (OOD) robustness, prediction consistency, resilience to adversaries, calibrated uncertainty estimates, and the ability to detect anomalous inputs. However, improving performance towards these goals is often a balancing act that today's methods cannot achieve without sacrificing performance on other safety axes. For instance, adversarial training improves adversarial robustness but sharply degrades other classifier performance metrics. Similarly, strong data augmentation and regularization techniques often improve OOD robustness but harm anomaly detection, raising the question of whether a Pareto improvement on all existing safety measures is possible. To meet this challenge, we design a new data augmentation strategy utilizing the natural structural complexity of pictures such as fractals, which outperforms numerous baselines, is near Pareto-optimal, and roundly improves safety measures.

Dell Zhang, Andre Freitas, Dacheng Tao et al.

This is the Proceedings of the AAAI-20 Workshop on Intelligent Process Automation (IPA-20) which took place in New York, NY, USA on February 7th 2020.

Daniel Huang, Prafulla Dhariwal, Dawn Song et al.

In this paper, we introduce a system called GamePad that can be used to explore the application of machine learning methods to theorem proving in the Coq proof assistant. Interactive theorem provers such as Coq enable users to construct machine-checkable proofs in a step-by-step manner. Hence, they provide an opportunity to explore theorem proving with human supervision. We use GamePad to synthesize proofs for a simple algebraic rewrite problem and train baseline models for a formalization of the Feit-Thompson theorem. We address position evaluation (i.e., predict the number of proof steps left) and tactic prediction (i.e., predict the next proof step) tasks, which arise naturally in tactic-based theorem proving.

Neil Zhenqiang Gong, Altay Ozen, Yu Wu et al.

Voice is envisioned to be a popular way for humans to interact with Internet-of-Things (IoT) devices. We propose a proximity-based user authentication method (called PIANO) for access control on such voice-powered IoT devices. PIANO leverages the built-in speaker, microphone, and Bluetooth that voice-powered IoT devices often already have. Specifically, we assume that a user carries a personal voice-powered device (e.g., smartphone, smartwatch, or smartglass), which serves as the user's identity. When another voice-powered IoT device of the user requires authentication, PIANO estimates the distance between the two devices by playing and detecting certain acoustic signals; PIANO grants access if the estimated distance is no larger than a user-selected threshold. We implemented a proof-of-concept prototype of PIANO. Through theoretical and empirical evaluations, we find that PIANO is secure, reliable, personalizable, and efficient.

Mario Frank, Tiffany Hwu, Sakshi Jain et al.

Martinovic et al. proposed a Brain-Computer-Interface (BCI) -based attack in which an adversary is able to infer private information about a user, such as their bank or area-of-living, by analyzing the user's brain activities. However, a key limitation of the above attack is that it is intrusive, requiring user cooperation, and is thus easily detectable and can be reported to other users. In this paper, we identify and analyze a more serious threat for users of BCI devices. We propose a it subliminal attack in which the victim is attacked at the levels below his cognitive perception. Our attack involves exposing the victim to visual stimuli for a duration of 13.3 milliseconds -- a duration usually not sufficient for conscious perception. The attacker analyzes subliminal brain activity in response to these short visual stimuli to infer private information about the user. If carried out carefully, for example by hiding the visual stimuli within screen content that the user expects to see, the attack may remain undetected. As a consequence, the attacker can scale it to many victims and expose them to the attack for a long time. We experimentally demonstrate the feasibility of our subliminal attack via a proof-of-concept study carried out with 27 subjects. We conducted experiments on users wearing Electroencephalography-based BCI devices, and used portrait pictures of people as visual stimuli which were embedded within the background of an innocuous video for a time duration not exceeding 13.3 milliseconds. Our experimental results show that it is feasible for an attacker to learn relevant private information about the user, such as whether the user knows the identity of the person for which the attacker is probing.

Mario Frank, Ralf Biedert, Eugene Ma et al.

We investigate whether a classifier can continuously authenticate users based on the way they interact with the touchscreen of a smart phone. We propose a set of 30 behavioral touch features that can be extracted from raw touchscreen logs and demonstrate that different users populate distinct subspaces of this feature space. In a systematic experiment designed to test how this behavioral pattern exhibits consistency over time, we collected touch data from users interacting with a smart phone using basic navigation maneuvers, i.e., up-down and left-right scrolling. We propose a classification framework that learns the touch behavior of a user during an enrollment phase and is able to accept or reject the current user by monitoring interaction with the touch screen. The classifier achieves a median equal error rate of 0% for intra-session authentication, 2%-3% for inter-session authentication and below 4% when the authentication test was carried out one week after the enrollment phase. While our experimental findings disqualify this method as a standalone authentication mechanism for long-term authentication, it could be implemented as a means to extend screen-lock time or as a part of a multi-modal biometric authentication system.