In previous research, there are plenty of arguments about textual backdoor evaluation, including diverse metrics and experiment settings. These valuable discussions motivate us to construct a rigorous benchmark and we highly appreciate their efforts. In this section, we briefly summarize existing opinions and provide a more detailed discussion on this topic. Table 9 summarizes the attackers OpenBackdoorimplements. Effectiveness Besides the mainstream ASR (also called LFR [20]) and CACC metrics, there are also other effectiveness metrics. Shen et al. [46] proposed to count the number of inserted triggers that can successfully flip the label. However, although inserting more triggers could benefit attack strength, the triggers also corrupt the sentences gradually, so it is also possible that the poisoned samples become "adversarial", and we can hardly distinguish. Shen et al. [45] also mentioned this issue, and they advised calculating the ASR difference between a poisoned model and a clean model as an effectiveness metric.

AI agents aim to solve complex tasks by combining text-based reasoning with external tool calls.Unfortunately, AI agents are vulnerable to prompt injection attacks where data returned by external tools hijacks the agent to execute malicious tasks.To measure the adversarial robustness of AI agents, we introduce AgentDojo, an evaluation framework for agents that execute tools over untrusted data.To capture the evolving nature of attacks and defenses, AgentDojo is not a static test suite, but rather an extensible environment for designing and evaluating new agent tasks, defenses, and adaptive attacks.We populate the environment with 97 realistic tasks (e.g., managing an email client, navigating an e-banking website, or making travel bookings), 629 security test cases, and various attack and defense paradigms from the literature.We find that AgentDojo poses a challenge for both attacks and defenses: state-of-the-art LLMs fail at many tasks (even in the absence of attacks), and existing prompt injection attacks break some security properties but not all. We hope that AgentDojo can foster research on new design principles for AI agents that solve common tasks in a reliable and robust manner.

Research on backdoor attacks in Federated Learning (FL) has accelerated in recent years, with new attacks and defenses continually proposed in an escalating arms race. However, the evaluation of these methods remains neither standardized nor reliable. First, there are severe inconsistencies in the evaluation settings across studies, and many rely on unrealistic threat models. Second, our code review uncovers semantic bugs in the official codebases of several attacks that artificially inflate their reported performance. These issues raise fundamental questions about whether current methods are truly effective or simply overfitted to narrow experimental setups. We introduce \textbf{BackFed}, a benchmark designed to standardize and stress-test FL backdoor evaluation by unifying attacks and defenses under a common evaluation framework that mirrors realistic FL deployments. Our benchmark on three representative datasets with three distinct architectures reveals critical limitations of existing methods. Malicious clients often require excessive training time and computation, making them vulnerable to server-enforced time constraints. Meanwhile, several defenses incur severe accuracy degradation or aggregation overhead. Popular defenses and attacks achieve limited performance in our benchmark, which challenges their previous efficacy claims. We establish BackFed as a rigorous and fair evaluation framework that enables more reliable progress in FL backdoor research.

In this paper, we propose a game-theoretic framework for studying attacks and defenses which exist in equilibrium.

Federated Learning (FL) is an emerging distributed machine learning paradigm enabling multiple clients to train a global model collaboratively without sharing their raw data. While FL enhances data privacy by design, it remains vulnerable to various security and privacy threats. This survey provides a comprehensive overview of more than 200 papers regarding the state-of-the-art attacks and defense mechanisms developed to address these challenges, categorizing them into security-enhancing and privacy-preserving techniques. Security-enhancing methods aim to improve FL robustness against malicious behaviors such as byzantine attacks, poisoning, and Sybil attacks. At the same time, privacy-preserving techniques focus on protecting sensitive data through cryptographic approaches, differential privacy, and secure aggregation. We critically analyze the strengths and limitations of existing methods, highlight the trade-offs between privacy, security, and model performance, and discuss the implications of non-IID data distributions on the effectiveness of these defenses. Furthermore, we identify open research challenges and future directions, including the need for scalable, adaptive, and energy-efficient solutions operating in dynamic and heterogeneous FL environments. Our survey aims to guide researchers and practitioners in developing robust and privacy-preserving FL systems, fostering advancements safeguarding collaborative learning frameworks' integrity and confidentiality.

As the social implementation of AI has been steadily progressing, research and development related to AI security has also been increasing. However, existing studies have been limited to organizing related techniques, attacks, defenses, and risks in terms of specific domains or AI elements. Thus, it extremely difficult to understand the relationships among them and how negative impacts on stakeholders are brought about. In this paper, we argue that the knowledge, technologies, and social impacts related to AI security should be holistically organized to help understand relationships among them. To this end, we first develop an AI security map that holistically organizes interrelationships among elements related to AI security as well as negative impacts on information systems and stakeholders. This map consists of the two aspects, namely the information system aspect (ISA) and the external influence aspect (EIA). The elements that AI should fulfill within information systems are classified under the ISA. The EIA includes elements that affect stakeholders as a result of AI being attacked or misused. For each element, corresponding negative impacts are identified. By referring to the AI security map, one can understand the potential negative impacts, along with their causes and countermeasures. Additionally, our map helps clarify how the negative impacts on AI-based systems relate to those on stakeholders. We show some findings newly obtained by referring to our map. We also provide several recommendations and open problems to guide future AI security communities.