Daeseon Choi

Hongjang Yang, Hyunsik Na, Daeseon Choi

LLM-based chatbot agents increasingly process user requests by combining natural-language reasoning with external tools such as web browsing. These capabilities improve usability, but they also create attack surfaces when untrusted external content is processed as part of a user' s task. This paper studies a privacy-leakage attack chain based on indirect prompt injection in black-box chatbot environments, where the attacker has no access to model weights, system prompts, or agent implementation details including how a trajectory is actually managed during its processing for a query. We first analyze how an attacker can hijack an agent' s intended task by crafting external content that appears benign to the victim while inducing the agent to execute an attacker-defined objective. We then evaluate a new prompt-injection technique, called exemplification, which uses a bridge in the external content to reframe the user prompt and the benign beginning of the retrieved page as few-shot examples before appending the attacker' s objective. We compare its attack success rate with a prior fake-completion technique. Finally, we demonstrate a proof-of-concept data-exfiltration chain using fictitious personal information in a controlled setting. Our results suggest that prompt injection, jailbreak-style instruction steering, and web-tool invocation can be combined into a feasible privacy-leakage path in deployed chatbot agents.

Seong-Gyu Park, Sohee Park, Jisu Lee et al.

Recent LLMs increasingly integrate reasoning mechanisms like Chain-of-Thought (CoT). However, this explicit reasoning exposes a new attack surface for inference-time backdoors, which inject malicious reasoning paths without altering model parameters. Because these attacks generate linguistically coherent paths, they effectively evade conventional detection. To address this, we propose STAR (State-Transition Amplification Ratio), a framework that detects backdoors by analyzing output probability shifts. STAR exploits the statistical discrepancy where a malicious input-induced path exhibits high posterior probability despite a low prior probability in the model's general knowledge. We quantify this state-transition amplification and employ the CUSUM algorithm to detect persistent anomalies. Experiments across diverse models (8B-70B) and five benchmark datasets demonstrate that STAR exhibits robust generalization capabilities, consistently achieving near-perfect performance (AUROC $\approx$ 1.0) with approximately $42\times$ greater efficiency than existing baselines. Furthermore, the framework proves robust against adaptive attacks attempting to bypass detection.

Heehwan Kim, Sungjune Park, Daeseon Choi

Large Language Models (LLMs) are generally equipped with guardrails to block the generation of harmful responses. However, existing defenses always assume that an external attacker crafts the harmful query, and the possibility of a model's own output becoming a new attack vector has not been sufficiently explored. In this study, we propose the Self-HarmLLM scenario, which uses a Mitigated Harmful Query (MHQ) generated by the same model as a new input. An MHQ is an ambiguous query whose original intent is preserved while its harmful nature is not directly exposed. We verified whether a jailbreak occurs when this MHQ is re-entered into a separate session of the same model. We conducted experiments on GPT-3.5-turbo, LLaMA3-8B-instruct, and DeepSeek-R1-Distill-Qwen-7B under Base, Zero-shot, and Few-shot conditions. The results showed up to 52% transformation success rate and up to 33% jailbreak success rate in the Zero-shot condition, and up to 65% transformation success rate and up to 41% jailbreak success rate in the Few-shot condition. By performing both prefix-based automated evaluation and human evaluation, we found that the automated evaluation consistently overestimated jailbreak success, with an average difference of 52%. This indicates that automated evaluation alone is not accurate for determining harmfulness. While this study is a toy-level study based on a limited query set and evaluators, it proves that our method can still be a valid attack scenario. These results suggest the need for a fundamental reconsideration of guardrail design and the establishment of a more robust evaluation methodology.

Sungjune Park, Heehwan Kim, Haehyun Cho et al.

This paper proposes a novel prompting approach, Judgment of Thought (JoT), specifically tailored for binary logical reasoning tasks. Despite advances in prompt engineering, existing approaches still face limitations in handling complex logical reasoning tasks. To address these issues, JoT introduces a multi-agent approach with three specialized roles$\unicode{x2010}$$\unicode{x2010}$$\unicode{x2010}$lawyer, prosecutor, and judge$\unicode{x2010}$$\unicode{x2010}$$\unicode{x2010}$where a high-level model acts as the judge, and lower-level models serve as lawyer and prosecutor to systematically debate and evaluate arguments. Experimental evaluations on benchmarks such as BigBenchHard and Winogrande demonstrate JoT's superior performance compared to existing prompting approaches, achieving notable improvements, including 98\% accuracy in Boolean expressions. Also, our ablation studies validate the critical contribution of each role, iterative refinement loops, and feedback mechanisms. Consequently, JoT significantly enhances accuracy, reliability, and consistency in binary reasoning tasks and shows potential for practical applications.

Hyunsik Na, Wonho Lee, Seungdeok Roh et al.

The advent of convenient and efficient fully unmanned stores equipped with artificial intelligence-based automated checkout systems marks a new era in retail. However, these systems have inherent artificial intelligence security vulnerabilities, which are exploited via adversarial patch attacks, particularly in physical environments. This study demonstrated that adversarial patches can severely disrupt object detection models used in unmanned stores, leading to issues such as theft, inventory discrepancies, and interference. We investigated three types of adversarial patch attacks -- Hiding, Creating, and Altering attacks -- and highlighted their effectiveness. We also introduce the novel color histogram similarity loss function by leveraging attacker knowledge of the color information of a target class object. Besides the traditional confusion-matrix-based attack success rate, we introduce a new bounding-boxes-based metric to analyze the practical impact of these attacks. Starting with attacks on object detection models trained on snack and fruit datasets in a digital environment, we evaluated the effectiveness of adversarial patches in a physical testbed that mimicked a real unmanned store with RGB cameras and realistic conditions. Furthermore, we assessed the robustness of these attacks in black-box scenarios, demonstrating that shadow attacks can enhance success rates of attacks even without direct access to model parameters. Our study underscores the necessity for robust defense strategies to protect unmanned stores from adversarial threats. Highlighting the limitations of the current defense mechanisms in real-time detection systems and discussing various proactive measures, we provide insights into improving the robustness of object detection models and fortifying unmanned retail environments against these attacks.

Andrew Yeo, Daeseon Choi

Large Language Models (LLMs) have seen rapid adoption in recent years, with industries increasingly relying on them to maintain a competitive advantage. These models excel at interpreting user instructions and generating human-like responses, leading to their integration across diverse domains, including consulting and information retrieval. However, their widespread deployment also introduces substantial security risks, most notably in the form of prompt injection and jailbreak attacks. To systematically evaluate LLM vulnerabilities -- particularly to external prompt injection -- we conducted a series of experiments on eight commercial models. Each model was tested without supplementary sanitization, relying solely on its built-in safeguards. The results exposed exploitable weaknesses and emphasized the need for stronger security measures. Four categories of attacks were examined: direct injection, indirect (external) injection, image-based injection, and prompt leakage. Comparative analysis indicated that Claude 3 demonstrated relatively greater robustness; nevertheless, empirical findings confirm that additional defenses, such as input normalization, remain necessary to achieve reliable protection.

Wonho Lee, Hyunsik Na, Jisu Lee et al.

The advent of adversarial patches poses a significant challenge to the robustness of AI models, particularly in the domain of computer vision tasks such as object detection. In contradistinction to traditional adversarial examples, these patches target specific regions of an image, resulting in the malfunction of AI models. This paper proposes Incremental Patch Generation (IPG), a method that generates adversarial patches up to 11.1 times more efficiently than existing approaches while maintaining comparable attack performance. The efficacy of IPG is demonstrated by experiments and ablation studies including YOLO's feature distribution visualization and adversarial training results, which show that it produces well-generalized patches that effectively cover a broader range of model vulnerabilities. Furthermore, IPG-generated datasets can serve as a robust knowledge foundation for constructing a robust model, enabling structured representation, advanced reasoning, and proactive defenses in AI security ecosystems. The findings of this study suggest that IPG has considerable potential for future utilization not only in adversarial patch defense but also in real-world applications such as autonomous vehicles, security systems, and medical imaging, where AI models must remain resilient to adversarial attacks in dynamic and high-stakes environments.

Seungmin Kim, Sohee Park, Donghyun Kim et al.

With the advancement of AI-based speech synthesis technologies such as Deep Voice, there is an increasing risk of voice spoofing attacks, including voice phishing and fake news, through unauthorized use of others' voices. Existing defenses that inject adversarial perturbations directly into audio signals have limited effectiveness, as these perturbations can easily be neutralized by speech enhancement methods. To overcome this limitation, we propose RoVo (Robust Voice), a novel proactive defense technique that injects adversarial perturbations into high-dimensional embedding vectors of audio signals, reconstructing them into protected speech. This approach effectively defends against speech synthesis attacks and also provides strong resistance to speech enhancement models, which represent a secondary attack threat. In extensive experiments, RoVo increased the Defense Success Rate (DSR) by over 70% compared to unprotected speech, across four state-of-the-art speech synthesis models. Specifically, RoVo achieved a DSR of 99.5% on a commercial speaker-verification API, effectively neutralizing speech synthesis attack. Moreover, RoVo's perturbations remained robust even under strong speech enhancement conditions, outperforming traditional methods. A user study confirmed that RoVo preserves both naturalness and usability of protected speech, highlighting its effectiveness in complex and evolving threat scenarios.