User interactions with personal assistants like Alexa, Google Home and Siri are typically initiated by a wake term or wakeword. Several personal assistants feature "follow-up" modes that allow users to make additional interactions without the need of a wakeword. For the system to only respond when appropriate, and to ignore speech not intended for it, utterances must be classified as device-directed or non-device-directed. State of the art systems have largely used acoustic features for this task, while others have used only lexical features or have added LM-based lexical features. We propose a directedness classifier that combines semantic lexical features with a lightweight acoustic feature and show it is effective in classifying directedness. The mixed-domain lexical and acoustic feature model is able to achieve 14% relative reduction of EER over a state of the art acoustic-only baseline model. Finally, we successfully apply transfer learning and semi-supervised learning to the model to improve accuracy even further.
Security in wireless ad-hoc networks is a complex issue. The wireless and dynamic nature of ad-hoc networks makes them more vulnerable to security attacks when compared with fixed networks. The existing routing protocols are optimized to perform the routing process without considering the security problem.. Black hole attack is one of the routing attacks in which, a malicious node uses the routing protocol to advertise itself as having the shortest path to the node whose packets it wants to intercept. In this paper we propose a certificate based authentication mechanism to counter the effect of black hole attack. Nodes authenticate each other by issuing certificates to neighboring nodes and generating public key without the need of any online centralized authority. The proposed scheme is implemented in two phases: certification phase and authentication phase following the route establishment process of On Demand Multicast Routing Protocol (ODMRP). The effectiveness of our mechanism is illustrated by simulations conducted using network simulator ns-2.
Information security has been a very active research over the last two decades. High connectivity and massive access to network resources combined with the widespread use of vulnerable software has resulted in hundreds of security breaches. An intruder may move between multiple nodes in the network to conceal the origin of attack. Distributed intrusion detection and prevention plays an increasingly important role in securing computer networks. To overcome the limitations of conventional intrusion detection systems, alerts are made in distributed intrusion detection system which are exchanged and correlated in a cooperative fashion. This paper presents an intelligent learning approach using Genetic Algorithm (GA) for distributed Intrusion Detection System (DIDS), which uses simple representation of rules and an effective fitness function. The proposed method is easier to implement while providing the flexibility. GA is used to select a subset of input features that increase the detection rate and decrease the false alarm rate. Also the generated rules can be used to detect the distributed network intrusions with effective and adaptive cost.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.