Related papers: The Little Phone That Could Ch-Ch-Chroot
As large language models (LLMs) are increasingly deployed as black-box components in real-world applications, red teaming has become essential for identifying potential risks. It tests LLMs with adversarial prompts to uncover…
In this paper, we investigate how to network smartphones for providing communications in disaster recovery. By bridging the gaps among different kinds of wireless networks, we have designed and implemented a system called TeamPhone, which…
Cybersecurity is a major challenge for large organizations. Traditional cybersecurity defense is reactive. Cybersecurity operations centers keep out adversaries and incident response teams clean up after break-ins. Recently a proactive…
We propose featured team automata to support variability in the development and analysis of teams, which are systems of reactive components that communicate according to specified synchronisation types. A featured team automaton concisely…
Chatbots, the common moniker for collaborative assistants, are Artificial Intelligence (AI) software that enables people to naturally interact with them to get tasks done. Although chatbots have been studied since the dawn of AI, they have…
Multi-turn interactions with language models (LMs) pose critical safety risks, as harmful intent can be strategically spread across exchanges. Yet, the vast majority of prior work has focused on single-turn safety, while adaptability and…
Trusted Computing is a security base technology that will perhaps be ubiquitous in a few years in personal computers and mobile devices alike. Despite its neutrality with respect to applications, it has raised some privacy concerns. We show…
A concern about cutting-edge or "frontier" AI foundation models is that an adversary may use the models for preparing chemical, biological, radiological, nuclear, (CBRN), cyber, or other attacks. At least two methods can identify foundation…
As large language models (LLMs) become increasingly capable, security and safety evaluation are crucial. While current red teaming approaches have made strides in assessing LLM vulnerabilities, they often rely heavily on human input and…
Formal methods have been largely thought of in the context of safety-critical systems, where they have achieved major acceptance. Tens of millions of people trust their lives every day to such systems, based on formal proofs rather than…
Text-to-image (T2I) models raise ethical and safety concerns due to their potential to generate inappropriate or harmful images. Evaluating these models' security through red-teaming is vital, yet white-box approaches are limited by their…
A digital security breach, by which confidential information is leaked, does not only affect the agent whose system is infiltrated, but is also detrimental to other agents socially connected to the infiltrated system. Although it has been…
Contact tracing is an effective tool in controlling the spread of infectious diseases such as COVID-19. It involves digital monitoring and recording of physical proximity between people over time with a central and trusted authority, so…
Large language models (LLMs) are susceptible to red teaming attacks, which can induce LLMs to generate harmful content. Previous research constructs attack prompts via manual or automatic methods, which have their own limitations on…
Covert communication allows us to transmit messages in such a way that it is not possible to detect that the communication is occurring. This provides protection in situations where knowledge that people are talking to each other may be…
When users leave their mobile devices unattended, or let others use them momentarily, they are susceptible to privacy breaches. Existing technological defenses, such as unlock authentication or account switching, have proven to be…
When building Large Language Models (LLMs), it is paramount to bear safety in mind and protect them with guardrails. Indeed, LLMs should never generate content promoting or normalizing harmful, illegal, or unethical behavior that may…
As the industry increasingly adopts agentic AI systems, understanding their unique vulnerabilities becomes critical. Prior research suggests that security flaws at the model level do not fully capture the risks present in agentic…
Systems operating in adversarial environments may inadvertently leak sensitive information to adversaries. To address this challenge, we revisit the linear-quadratic control framework and introduce deception to actively mislead adversaries.…
Cooperation is a fundamental human concept studied across multiple social and biological disciplines. Traditional methods for eliciting an individual's propensity to cooperate have included surveys and laboratory experiments and multiple…