There are many research papers devoted to the state identification problem of finite state machines (FSMs) which are widely used for analysis of discrete event systems. A deterministic complete reduced FSM always has ...
详细信息
The article is devoted to the development of means for recognition of the emotions of the speaker, based on the neural network analysis of fixed fragments of the voice signal. The possibility of improving recognition ...
详细信息
Nowadays in Ukraine, there is a need for automation and removal of human from the processes of decision making on civil law regulation of contract for the provision of in vitro fertilization, which can significantly i...
详细信息
The purpose of this study is increasing the usability of the user interfaces (UI) by ensuring their compliance with Gestalt principles. The developed method of evaluating the compliance of the UI with Gestalt principl...
详细信息
The details of future failures of systems that are serially produced are important for the remanufacturing departments to frame an optimal strategy for the post series supply. The existing one-dimensional and two-dime...
详细信息
Remanufacturing, a popular term in the automotive industry, helps to reduce the need for raw materials, increase profitability, is an effective solution to the challenges of postal items and also benefits the environm...
详细信息
To remanufacture automobile systems, it is important to understand the future failure rate of the serially produced systems. In addition, remanufacturing departments need the information about the number of cores to r...
详细信息
Nowadays the information of future rates of systems produced in series appears to be crucial to the production plan, especially for remanufacturing departments. Several questions need to be answered: the number of cor...
详细信息
— Cyber Physical systems (CPS) are predestined for use in Industry 4.0 applications. However, the interaction between the virtual and physical world also creates risks that is essential to be controlled. In highly au...
详细信息
Data poisoning attacks, where adversaries manipulate training data to degrade model performance, are an emerging threat as machine learning becomes widely deployed in sensitive applications. This paper provides a comp...
Data poisoning attacks, where adversaries manipulate training data to degrade model performance, are an emerging threat as machine learning becomes widely deployed in sensitive applications. This paper provides a comprehensive overview of data poisoning including attack techniques, adversary incentives, impacts on security and reliability, detection methods, defenses, and key research gaps. We examine label flipping, instance injection, backdoors, and other attack categories that enable malicious outcomes ranging from IP theft to accidents in autonomous systems. Promising detection approaches include statistical tests, robust learning, and forensics. However, significant challenges remain in translating academic defenses like adversarial training and sanitization into practical tools ready for operational use. With safety and trustworthiness at stake, more research on benchmarking evaluations, adaptive attacks, fundamental tradeoffs, and real-world deployment of defenses is urgently needed. Understanding vulnerabilities and developing resilient machine learning pipelines will only grow in importance as data integrity is fundamental to developing safe artificial intelligence.
暂无评论