Long-range attacks pose a significant threat to the integrity of Proof-of-Stake (PoS) blockchains by enabling adversaries to reconstruct an alternative chain history embedded with fraudulent transactions. These attacks can deceive honest participants into accepting a maliciously crafted branch as the canonical chain. While Key Evolving Signature (KES) schemes are widely adopted to mitigate such threats, they typically rely on the assumption that validators behave honestly. In this work, we challenge this assumption by demonstrating how a malicious validator can exploit inherent limitations in existing KES-based mechanisms to mount a successful long-range attack. To address this critical vulnerability, we introduce a novel cryptographic construction that combines one-time signatures with commitment schemes. Our approach imposes constraints on the signing capabilities of validators, thereby significantly reducing the feasibility of long-range attacks. We provide rigorous formal security proofs to substantiate the robustness of our scheme and conduct a comprehensive performance evaluation. The results show that our solution is both computationally and storage efficient, making it a practical and scalable defense mechanism for real-world PoS blockchain deployments.

The advent of large language models (LLMs) has revolutionized the field of text generation, producing outputs that closely mimic human-like writing. Although academic and industrial institutions have developed detectors to prevent the malicious usage of LLM-generated texts, other research has doubt about the robustness of these systems. To stress test these detectors, we introduce a humanized proxy-attack (HUMPA) strategy that effortlessly compromises LLMs, causing them to produce outputs that align with human-written text and mislead detection systems. Our method attacks the source model by leveraging a reinforcement learning (RL) fine-tuned humanized small language model (SLM) in the decoding phase. Through an in-depth analysis, we demonstrate that our attack strategy is capable of generating responses that are indistinguishable to detectors, preventing them from differentiating between machine-generated and human-written text. We conduct systematic evaluations on extensive datasets using proxy-attacked open-source models, including Llama2-13B, Llama3-70B, and Mixtral-8x7B in both white- and black-box settings. Our findings show that the proxy-attack strategy effectively deceives the leading detectors, resulting in an average AUROC drop of 70.4% across multiple datasets, with a maximum drop of 95.0% on a single dataset. Furthermore, in cross-discipline scenarios, our strategy also bypasses these detectors, leading to a significant relative decrease of up to 90.9%, while in cross-language scenario, the drop reaches 91.3%. Despite our proxy-attack strategy successfully bypassing the detectors with such significant relative drops, we find that the generation quality of the attacked models remains preserved, even within a modest utility budget, when compared to the text produced by the original, unattacked source model.

Recent work has revealed that large language models (LLMs) can exhibit emergent theory-of-mind (ToM) capabilities—inferring human beliefs, desires, and intentions from text alone. Yet, everyday social reasoning often unfolds visually in dynamic contexts. This paper investigates whether multimodal LLMs can similarly demonstrate ToM skills in video-based tasks. Concretely, we propose a pipeline that fuses video and text signals, retrieves the most relevant frames for each query, and answers questions requiring spatio-temporal social understanding. We introduce a new frame localization benchmark, Theory of Mind Localization (ToMLoc), and show that finetuning a state-of-the-art Video-ChatGPT model on ToMLoc significantly improves performance on Social-Iq 2.0. Our results suggest that bridging textual and visual modalities is essential for capturing complex mental states in real-world scenarios. Moreover, retrieving key frames enhances interpretability by revealing how the model arrives at its inferences. These findings highlight the promise of video-based approaches for achieving more human-like social intelligence in LLMs.

Scene graph generation (SGG) models seek to detect relationships between objects in a given image. One challenge in this area is the biased distribution of predicates in the dataset and the semantic space. Recent works incorporating knowledge graphs with scene graphs prove effective in improving recall for the tail predicate classes. Moreover, many recent SGG approaches with promising results explicitly redistribute the predicates in both the training process and in the prediction step. To incorporate external knowledge, we construct a commonsense knowledge graph by integrating ConceptNet and Wikidata. To explicitly unbias SGG with knowledge in the reasoning process, we propose a novel framework, Explicit Ontological Adjustment (EOA), to adjust the graph model predictions with knowledge priors. We use the edge matrix from the commonsense knowledge graph as a module in the graph neural network model to refine the relationship detection process. This module proves effective in alleviating the long-tail distribution of predicates. When combined, we show that these modules achieve state-of-the-art performance on the Visual Genome dataset in most cases. The source code is available at https://github.com/zhanwenchen/eoa.

Summary While the social network services (SNS) have dominate the ways that people communicate with each other on the Internet, identity impersonation remains to be a serious issue that needs to be solved due to the anonymity in the cyber network. Currently, the potential solution to the problem relies heavily on the administration from the central server, which requires intensive workload of the identity management. In this article, we propose a threshold identity authentication signature scheme to solve the impersonation problem from the protocol layer rather than software design in the traditional upper level. In our scheme, with the help of some authenticated accounts, trusted relationship can be shared in a group to other unauthenticated accounts, which largely decrease the workload of authenticating all the accounts. Users are given the ability to verify other accounts' identity information by their signatures. Then, we establish three security goals to prevent the malicious adversary to launch the impersonation attack on a group. We claim that our scheme is suitable for the SNS scenario since the procedure of generating a signature to prove the identity requires little computation cost, it is user‐friendly especially on the lightweight devices such as mobile devices and so on.