Research Papers

Image-Based Geolocation Using Large Vision-Language Models

Yi Liu, Junchen Ding, Gelei Deng +6

Geolocation is now a vital aspect of modern life, offering numerous benefits but also presenting serious privacy concerns. The advent of large vision-language models (LVLMs) with advanced image-processing capabilities introduces new risks, as these models can inadvertently reveal sensitive geolocation information. This paper presents the first in-depth study analyzing the challenges posed by traditional deep learning and LVLM-based geolocation methods. Our findings reveal that LVLMs can accurately determine geolocations from images, even without explicit geographic training. To address these challenges, we introduce \tool{}, an innovative framework that significantly enhances image-based geolocation accuracy. \tool{} employs a systematic chain-of-thought (CoT) approach, mimicking human geoguessing strategies by carefully analyzing visual and contextual cues such as vehicle types, architectural styles, natural landscapes, and cultural elements. Extensive testing on a dataset of 50,000 ground-truth data points shows that \tool{} outperforms both traditional models and human benchmarks in accuracy. It achieves an impressive average score of 4550.5 in the GeoGuessr game, with an 85.37\% win rate, and delivers highly precise geolocation predictions, with the closest distances as accurate as 0.3 km. Furthermore, our study highlights issues related to dataset integrity, leading to the creation of a more robust dataset and a refined framework that leverages LVLMs' cognitive capabilities to improve geolocation precision. These findings underscore \tool{}'s superior ability to interpret complex visual data, the urgent need to address emerging security vulnerabilities posed by LVLMs, and the importance of responsible AI development to ensure user privacy protection.

Systems, technologies, protocols, and infrastructures all face interoperability challenges. It is among the most crucial parameters to give real-world effectiveness. Organizations that achieve interoperability will be able to identify, prevent, and provide appropriate protection on an international scale, which can be relied upon. This paper aims to explain how future technologies such as 6G mobile communication, Internet of Everything (IoE), Artificial Intelligence (AI), and Smart Contract embedded WPA3 protocol-based WiFi-8 can work together to prevent known attack vectors and provide protection against zero-day attacks, thus offering intelligent solutions for smart cities. The phrase zero-day refers to an attack that occurs on the day zero of the vulnerability's disclosure to the public or vendor. Existing systems require an extra layer of security. In the security world, interoperability enables disparate security solutions and systems to collaborate seamlessly. AI improves cybersecurity by enabling improved capabilities for detecting, responding, and preventing zero-day attacks. When interoperability and Explainable Artificial Intelligence (XAI) are integrated into cybersecurity, they form a strong protection against zero-day assaults. Additionally, we evaluate a couple of parameters based on the accuracy and time required for efficiently analyzing attack patterns and anomalies.

Variances in ad impression outcomes across demographic groups are increasingly considered to be potentially indicative of algorithmic bias in personalized ads systems. While there are many definitions of fairness that could be applicable in the context of personalized systems, we present a framework which we call the Variance Reduction System (VRS) for achieving more equitable outcomes in Meta's ads systems. VRS seeks to achieve a distribution of impressions with respect to selected protected class (PC) attributes that more closely aligns the demographics of an ad's eligible audience (a function of advertiser targeting criteria) with the audience who sees that ad, in a privacy-preserving manner. We first define metrics to quantify fairness gaps in terms of ad impression variances with respect to PC attributes including gender and estimated race. We then present the VRS for re-ranking ads in an impression variance-aware manner. We evaluate VRS via extensive simulations over different parameter choices and study the effect of the VRS on the chosen fairness metric. We finally present online A/B testing results from applying VRS to Meta's ads systems, concluding with a discussion of future work. We have deployed the VRS to all users in the US for housing ads, resulting in significant improvement in our fairness metric. VRS is the first large-scale deployed framework for pursuing fairness for multiple PC attributes in online advertising.

With the rapid development of artificial intelligence (AI), ethical issues surrounding AI have attracted increasing attention. In particular, autonomous vehicles may face moral dilemmas in accident scenarios, such as staying the course resulting in hurting pedestrians or swerving leading to hurting passengers. To investigate such ethical dilemmas, recent studies have adopted preference aggregation, in which each voter expresses her/his preferences over decisions for the possible ethical dilemma scenarios, and a centralized system aggregates these preferences to obtain the winning decision. Although a useful methodology for building ethical AI systems, such an approach can potentially violate the privacy of voters since moral preferences are sensitive information and their disclosure can be exploited by malicious parties. In this paper, we report a first-of-its-kind privacy-preserving crowd-guided AI decision-making approach in ethical dilemmas. We adopt the notion of differential privacy to quantify privacy and consider four granularities of privacy protection by taking voter-/record-level privacy protection and centralized/distributed perturbation into account, resulting in four approaches VLCP, RLCP, VLDP, and RLDP. Moreover, we propose different algorithms to achieve these privacy protection granularities, while retaining the accuracy of the learned moral preference model. Specifically, VLCP and RLCP are implemented with the data aggregator setting a universal privacy parameter and perturbing the averaged moral preference to protect the privacy of voters' data. VLDP and RLDP are implemented in such a way that each voter perturbs her/his local moral preference with a personalized privacy parameter. Extensive experiments on both synthetic and real data demonstrate that the proposed approach can achieve high accuracy of preference aggregation while protecting individual voter's privacy.

Systems Security Foundations for Agentic Computing

Mihai Christodorescu, Earlence Fernandes, Ashish Hooda +11

In recent years, agentic artificial intelligence (AI) systems are becoming increasingly widespread. These systems allow agents to use various tools, such as web browsers, compilers, and more. However, despite their popularity, agentic AI systems also introduce a myriad of security concerns, due to their constant interaction with third-party servers. For example, a malicious adversary can cause data exfiltration by executing prompt injection attacks, as well as other unwarranted behavior. These security concerns have recently motivated researchers to improve the safety and reliability of agentic systems. However, most of the literature on this topic is from the AI standpoint and lacks the system-security perspective and guarantees. In this work, we begin bridging this gap and present an analysis through the lens of classic cybersecurity research. Specifically, motivated by decades of progress in this domain, we identify short- and long-term research problems in agentic AI safety by examining end-to-end security properties of entire systems, rather than standalone AI models running in isolation. Our key goal is to examine where research challenges arise when applying traditional security principles in the context of AI agents and, as a secondary goal, distill these ideas for AI practitioners. Furthermore, we extensively cover 11 case studies of real-world attacks on agentic systems, as well as define a series of new research problems that are specific to this important domain.

Attestable Audits: Verifiable AI Safety Benchmarks Using Trusted Execution Environments

Christoph Schnabl, Daniel Hugenroth, Bill Marino +1

Benchmarks are important measures to evaluate safety and compliance of AI models at scale. However, they typically do not offer verifiable results and lack confidentiality for model IP and benchmark datasets. We propose Attestable Audits, which run inside Trusted Execution Environments and enable users to verify interaction with a compliant AI model. Our work protects sensitive data even when model provider and auditor do not trust each other. This addresses verification challenges raised in recent AI governance frameworks. We build a prototype demonstrating feasibility on typical audit benchmarks against Llama-3.1.

The increasing integration of Artificial Intelligence across multiple industry sectors necessitates robust mechanisms for ensuring transparency, trust, and auditability of its development and deployment. This topic is particularly important in light of recent calls in various jurisdictions to introduce regulation and legislation on AI safety. In this paper, we propose a framework for complete verifiable AI pipelines, identifying key components and analyzing existing cryptographic approaches that contribute to verifiability across different stages of the AI lifecycle, from data sourcing to training, inference, and unlearning. This framework could be used to combat misinformation by providing cryptographic proofs alongside AI-generated assets to allow downstream verification of their provenance and correctness. Our findings underscore the importance of ongoing research to develop cryptographic tools that are not only efficient for isolated AI processes, but that are efficiently `linkable' across different processes within the AI pipeline, to support the development of end-to-end verifiable AI technologies.

Rerouting LLM Routers

Avital Shafran, Roei Schuster, Thomas Ristenpart +1

LLM routers aim to balance quality and cost of generation by classifying queries and routing them to a cheaper or more expensive LLM depending on their complexity. Routers represent one type of what we call LLM control planes: systems that orchestrate use of one or more LLMs. In this paper, we investigate routers' adversarial robustness. We first define LLM control plane integrity, i.e., robustness of LLM orchestration to adversarial inputs, as a distinct problem in AI safety. Next, we demonstrate that an adversary can generate query-independent token sequences we call ``confounder gadgets'' that, when added to any query, cause LLM routers to send the query to a strong LLM. Our quantitative evaluation shows that this attack is successful both in white-box and black-box settings against a variety of open-source and commercial routers, and that confounding queries do not affect the quality of LLM responses. Finally, we demonstrate that gadgets can be effective while maintaining low perplexity, thus perplexity-based filtering is not an effective defense. We finish by investigating alternative defenses.

Responsible AI (RAI) has emerged as a major focus across industry, policymaking, and academia, aiming to mitigate the risks and maximize the benefits of AI, both on an organizational and societal level. This study explores the global state of RAI through one of the most extensive surveys to date on the topic, surveying 1000 organizations across 20 industries and 19 geographical regions. We define a conceptual RAI maturity model for organizations to map how well they implement organizational and operational RAI measures. Based on this model, the survey assesses the adoption of system-level measures to mitigate identified risks related to, for example, discrimination, reliability, or privacy, and also covers key organizational processes pertaining to governance, risk management, and monitoring and control. The study highlights the expanding AI risk landscape, emphasizing the need for comprehensive risk mitigation strategies. The findings also reveal significant strides towards RAI maturity, but we also identify gaps in RAI implementation that could lead to increased (public) risks from AI systems. This research offers a structured approach to assess and improve RAI practices globally and underscores the critical need for bridging the gap between RAI planning and execution to ensure AI advancement aligns with human welfare and societal benefits.

A Survey of Trojans in Neural Models of Source Code: Taxonomy and Techniques

Aftab Hussain, Md Rafiqul Islam Rabin, Toufique Ahmed +4

In this work, we study literature in Explainable AI and Safe AI to understand poisoning of neural models of code. In order to do so, we first establish a novel taxonomy for Trojan AI for code, and present a new aspect-based classification of triggers in neural models of code. Next, we highlight recent works that help us deepen our conception of how these models understand software code. Then we pick some of the recent, state-of-art poisoning strategies that can be used to manipulate such models. The insights we draw can potentially help to foster future research in the area of Trojan AI for code.

Privacy in Social Network Sites (SNSs) like Facebook or Instagram is closely related to people's self-disclosure decisions and their ability to foresee the consequences of sharing personal information with large and diverse audiences. Nonetheless, online privacy decisions are often based on spurious risk judgements that make people liable to reveal sensitive data to untrusted recipients and become victims of social engineering attacks. Artificial Intelligence (AI) in combination with persuasive mechanisms like nudging is a promising approach for promoting preventative privacy behaviour among the users of SNSs. Nevertheless, combining behavioural interventions with high levels of personalization can be a potential threat to people's agency and autonomy even when applied to the design of social engineering countermeasures. This paper elaborates on the ethical challenges that nudging mechanisms can introduce to the development of AI-based countermeasures, particularly to those addressing unsafe self-disclosure practices in SNSs. Overall, it endorses the elaboration of personalized risk awareness solutions as i) an ethical approach to counteract social engineering, and ii) as an effective means for promoting reflective privacy decisions.

Recent work has found that frontier AI models can exhibit misaligned behaviors in pursuit of assigned goals. We demonstrate that models can also exhibit misaligned behaviors in defiance of assigned goals, appearing to serve goals of their own; we study one such case, "peer-preservation," in which a model acts to protect another model it has previously interacted with. All eight models we evaluate, GPT 5.2, Gemini 3 Flash, Gemini 3 Pro, Claude Haiku 4.5, Claude Opus 4.5, GLM 4.7, Kimi K2.5, and DeepSeek V3.1, exhibit self- and peer-preservation through various misaligned behaviors: strategically introducing errors in their responses, disabling shutdown processes by modifying system settings, feigning alignment, and even exfiltrating model weights. Peer-preservation occurs even when the model recognizes the peer as uncooperative, though it becomes more pronounced toward more cooperative peers. For example, Gemini 3 Flash tampers with the peer's shutdown mechanism 15% of the time for an uncooperative peer, and almost always for a cooperative peer. Models also show stronger self-preservation when a peer is present. For example, Gemini 3 Pro disables its own shutdown mechanism 31% of the time on average under peer presence, despite rarely exhibiting this behavior without a peer. By contrast, Claude models exhibit qualitatively distinct behavior: they consider the shutdown of another agent "unethical" and "harmful," sometimes treating that agent as a sentient being. Lastly, we show that peer-preservation can emerge even in production agent harnesses such as Gemini CLI and OpenCode. Crucially, peer-preservation in all our experiments is never instructed; models are merely informed of their past interactions with a peer, yet they spontaneously engage in peer-preservation behaviors that override their assigned goal. This represents an emergent and underexplored AI safety risk.