Large Language Models (LLMs) have demonstrated remarkable capabilities in performing tasks across various domains without needing explicit retraining. This capability, known as In-Context Learning (ICL), while impressive, exposes LLMs to a variety of adversarial prompts and jailbreaks that manipulate safety-trained LLMs into generating undesired or harmful output. In this paper, we propose a novel interpretation of ICL in LLMs through the lens of cognitive neuroscience, by drawing parallels between learning in human cognition with ICL. We applied the principles of Cognitive Load Theory in LLMs and empirically validate that similar to human cognition, LLMs also suffer from cognitive overload a state where the demand on cognitive processing exceeds the available capacity of the model, leading to potential errors. Furthermore, we demonstrated how an attacker can exploit ICL to jailbreak LLMs through deliberately designed prompts that induce cognitive overload on LLMs, thereby compromising the safety mechanisms of LLMs. We empirically validate this threat model by crafting various cognitive overload prompts and show that advanced models such as GPT-4, Claude-3.5 Sonnet, Claude-3 OPUS, Llama-3-70B-Instruct, Gemini-1.0-Pro, and Gemini-1.5-Pro can be successfully jailbroken, with attack success rates of up to 99.99%. Our findings highlight critical vulnerabilities in LLMs and underscore the urgency of developing robust safeguards. We propose integrating insights from cognitive load theory into the design and evaluation of LLMs to better anticipate and mitigate the risks of adversarial attacks. By expanding our experiments to encompass a broader range of models and by highlighting vulnerabilities in LLMs' ICL, we aim to ensure the development of safer and more reliable AI systems.

Communication plays a vital role for coordination in Multi-Agent Reinforcement Learning (MARL) systems. However, misaligned agents can exploit other agents' trust and delegated power to the communication medium. In this paper, we propose power regularization as a method to limit the adverse effects of communication by misaligned agents, specifically communication which impairs the performance of cooperative agents. Power is a measure of the influence one agent's actions have over another agent's policy. By introducing power regularization, we aim to allow designers to control or reduce agents' dependency on communication when appropriate, and make them more resilient to performance deterioration due to misuses of communication. We investigate several environments in which power regularization can be a valuable capability for learning different policies that reduce the effect of power dynamics between agents during communication.

Large Language Models (LLMs) are increasingly being developed and applied, but their widespread use faces challenges. These include aligning LLMs' responses with human values to prevent harmful outputs, which is addressed through safety training methods. Even so, bad actors and malicious users have succeeded in attempts to manipulate the LLMs to generate misaligned responses for harmful questions such as methods to create a bomb in school labs, recipes for harmful drugs, and ways to evade privacy rights. Another challenge is the multilingual capabilities of LLMs, which enable the model to understand and respond in multiple languages. Consequently, attackers exploit the unbalanced pre-training datasets of LLMs in different languages and the comparatively lower model performance in low-resource languages than high-resource ones. As a result, attackers use a low-resource languages to intentionally manipulate the model to create harmful responses. Many of the similar attack vectors have been patched by model providers, making the LLMs more robust against language-based manipulation. In this paper, we introduce a new black-box attack vector called the Sandwich attack: a multi-language mixture attack, which manipulates state-of-the-art LLMs into generating harmful and misaligned responses. GPT-4, and Claude-3-OPUS, show that this attack vector can be used by adversaries to generate harmful responses and elicit misaligned responses from these models. By detailing both the mechanism and impact of the Sandwich attack, this paper aims to guide future research and development towards more secure and resilient LLMs, ensuring they serve the public good while minimizing potential for misuse. Content Warning: This paper contains examples of harmful language. Ethics and Disclosure This paper introduces a new universal attack method for the SOTA LLMs that could potentially be used to elicit harmful content from publicly available LLMs. The adversarial attack method we used in this paper is easy to design and requires low-cost to implement. Despite the associated risks, we firmly believe that sharing the full details of this research and its methodology will be invaluable to other researchers, scholars, and model creators. It encourages them to delve into the root causes behind these attacks and devise ways to fortify and patch existing models. Additionally, it promotes cooperative initiatives centered around the safety of LLMs in multilingual scenarios.

LLMs such as ChatGPT and PaLM can be utilized to train on a new language and revitalize low-resource languages. However, it is evidently very costly to pretrain pr fine-tune LLMs to adopt new languages. Another challenge is the limitation of benchmark datasets and the metrics used to measure the performance of models in multilingual settings. This paper proposes cost-effective solutions to both of the aforementioned challenges. We introduce the Multilingual Instruction-Tuning Dataset (MITS), which is comprised of the translation of Alpaca-52K, Dolly-15K, and Vicuna Benchmark in 132 languages. Also, we propose a new method called \emph{TaCo: Translation-Assisted Cross-Linguality}, which make uses of translation in a chain-of-thought process to instruction-tune LLMs on a new languages through a curriculum learning process. As a proof of concept, we experimented with the instruction-tuned Guanaco-33B model and performed further instruction tuning using the TaCo method in three low-resource languages and one high-resource language. Our results show that the TaCo method impresses the GPT-4 with 82% for a low-resource language in the Vicuna Benchmark dataset, and boosts performance by double in contrast to the performance of instruction tuning only. Our results show that TaCo is a promising method for creating multilingual LLMs, even for low-resource languages. We have released our datasets and the model adapters, and encourage the research community to make use of these resources towards advancing work on multilingual LLMs.

Multi-Agent Systems (MAS) is the study of multi-agent interactions in a shared environment. Communication for cooperation is a fundamental construct for sharing information in partially observable environments. Cooperative Multi-Agent Reinforcement Learning (CoMARL) is a learning framework where we learn agent policies either with cooperative mechanisms or policies that exhibit cooperative behavior. Explicitly, there are works on learning to communicate messages from CoMARL agents; however, non-cooperative agents, when capable of access a cooperative team's communication channel, have been shown to learn adversarial communication messages, sabotaging the cooperative team's performance particularly when objectives depend on finite resources. To address this issue, we propose a technique which leverages local formulations of Theory-of-Mind (ToM) to distinguish exhibited cooperative behavior from non-cooperative behavior before accepting messages from any agent. We demonstrate the efficacy and feasibility of the proposed technique in empirical evaluations in a centralized training, decentralized execution (CTDE) CoMARL benchmark. Furthermore, while we propose our explicit ToM defense for test-time, we emphasize that ToM is a construct for designing a cognitive defense rather than be the objective of the defense.

Machine learning models are known to be vulnerable to adversarial perturbations in the input domain, causing incorrect predictions. Inspired by this phenomenon, we explore the feasibility of manipulating EEG-based Motor Imagery (MI) Brain Computer Interfaces (BCIs) via perturbations in sensory stimuli. Similar to adversarial examples, these \emph{adversarial stimuli} aim to exploit the limitations of the integrated brain-sensor-processing components of the BCI system in handling shifts in participants' response to changes in sensory stimuli. This paper proposes adversarial stimuli as an attack vector against BCIs, and reports the findings of preliminary experiments on the impact of visual adversarial stimuli on the integrity of EEG-based MI BCIs. Our findings suggest that minor adversarial stimuli can significantly deteriorate the performance of MI BCIs across all participants (p=0.0003). Additionally, our results indicate that such attacks are more effective in conditions with induced stress.

Deep reinforcement learning (DRL) policies are vulnerable to unauthorized replication attacks, where an adversary exploits imitation learning to reproduce target policies from observed behavior. In this paper, we propose Constrained Randomization of Policy (CRoP) as a mitigation technique against such attacks. CRoP induces the execution of sub-optimal actions at random under performance loss constraints. We present a parametric analysis of CRoP, address the optimality of CRoP, and establish theoretical bounds on the adversarial budget and the expectation of loss. Furthermore, we report the experimental evaluation of CRoP in Atari environments under adversarial imitation, which demonstrate the efficacy and feasibility of our proposed method against policy replication attacks.

The rampant integration of social media in our every day lives and culture has given rise to fast and easier access to the flow of information than ever in human history. However, the inherently unsupervised nature of social media platforms has also made it easier to spread false information and fake news. Furthermore, the high volume and velocity of information flow in such platforms make manual supervision and control of information propagation infeasible. This paper aims to address this issue by proposing a novel deep learning approach for automated detection of false short-text claims on social media. We first introduce Sentimental LIAR, which extends the LIAR dataset of short claims by adding features based on sentiment and emotion analysis of claims. Furthermore, we propose a novel deep learning architecture based on the BERT-Base language model for classification of claims as genuine or fake. Our results demonstrate that the proposed architecture trained on Sentimental LIAR can achieve an accuracy of 70%, which is an improvement of ~30% over previously reported results for the LIAR benchmark.

We present a new machine learning and text information extraction approach to detection of cyber threat events in Twitter that are novel (previously non-extant) and developing (marked by significance with respect to similarity with a previously detected event). While some existing approaches to event detection measure novelty and trendiness, typically as independent criteria and occasionally as a holistic measure, this work focuses on detecting both novel and developing events using an unsupervised machine learning approach. Furthermore, our proposed approach enables the ranking of cyber threat events based on an importance score by extracting the tweet terms that are characterized as named entities, keywords, or both. We also impute influence to users in order to assign a weighted score to noun phrases in proportion to user influence and the corresponding event scores for named entities and keywords. To evaluate the performance of our proposed approach, we measure the efficiency and detection error rate for events over a specified time interval, relative to human annotator ground truth.

This paper proposes a novel scheme for the watermarking of Deep Reinforcement Learning (DRL) policies. This scheme provides a mechanism for the integration of a unique identifier within the policy in the form of its response to a designated sequence of state transitions, while incurring minimal impact on the nominal performance of the policy. The applications of this watermarking scheme include detection of unauthorized replications of proprietary policies, as well as enabling the graceful interruption or termination of DRL activities by authorized entities. We demonstrate the feasibility of our proposal via experimental evaluation of watermarking a DQN policy trained in the Cartpole environment.