Search

Devices such as cell phones, vehicles, and smart home systems often generate time-series data that are crucial for developing effective AI applications. However, directly analyzing and using these local data could violate user privacy. Although synthetic data offer a promising alternative to real data, current methods for synthesizing time-series data either require servers to first collect large amounts of real user data or fail to offer adequate privacy protection for training generative models. In order to address the limitations in existing solutions, we introduce FedSTDG, an innovative framework for the distributed synthesis of time-series data under strict local differential privacy (LDP) guarantees. Our approach combines federated learning (FL) with a recurrent Wasserstein autoencoder (RWAE) to train generative models without the need to collect real local data directly. Additionally, we propose an enhanced LDP-FL mechanism that incorporates a private dimension selection algorithm and an adaptive learning rate adjustment algorithm. Our approach substantially reduces privacy risks during FL training and improves the model utility compared to existing LDP-FL methods. Extensive experiments on open-source datasets show that synthetic data generated by our framework can reduce downstream prediction errors by 20% to 85% compared to baseline models under equivalent privacy constraints. Furthermore, the framework offers stronger privacy protection and robustness compared to traditional nonprivate FL mechanisms.

Cybersecurity has become a key factor that determines the success or failure of companies that rely on information systems. Therefore, investment in cybersecurity is an important financial and operational decision. Typical information technology investments aim to create value, whereas cybersecurity investments aim to minimize loss incurred by cyber attacks. Admittedly, cybersecurity investment has become an increasingly complex one, since information systems are typically subject to frequent attacks, whose arrival and impact fluctuate stochastically. Furthermore, cybersecurity measures and improvements, such as patches, become available at random points in time making investment decisions even more challenging. We propose and develop an analytical real options framework that incorporates major components relevant to cybersecurity practice, and analyze how optimal cybersecurity investment decisions perform for a private firm. The novelty of this paper is that it provides analytical solutions that lend themselves to intuitive interpretations regarding the effect of timing and cybersecurity risk on investment behavior using real options theory. Such aspects are frequently not implemented within economic models that support policy initiatives. However, if these are not properly understood, security controls will not be properly set resulting in a dynamic inefficiency reflected in cycles of over or under investment, and, in turn, increased cybersecurity risk following corrective policy actions. Results indicate that greater uncertainty over the cost of cybersecurity attacks raises the value of an embedded option to invest in cybersecurity. This increases the incentive to suspend operations temporarily in order to install a cybersecurity patch that will make the firm more resilient to cybersecurity breaches. Similarly, greater likelihood associated with the availability of a cybersecurity patch increases the value of the option to invest in cybersecurity. However, the absence of an embedded investment option increases the incentive to delay the permanent abandonment of the company's operation due to the irreversible nature of the decision.