The distributed detection of botnets may induce heavy computation and communication costs to network devices. Each device in related scheme only has a regional view of Internet, so it is hard to detect botnet comprehe...The distributed detection of botnets may induce heavy computation and communication costs to network devices. Each device in related scheme only has a regional view of Internet, so it is hard to detect botnet comprehensively. In this paper, we propose a lightweight real-time botnet detection framework called Bot-Guard, which uses the global landscape and flexible configurability of software defined network （SDN） to identify botnets promptly. SDN, as a new network framework, can make centralized control in botnet detection, but there are still some challenges in such detections. We give a convex lens imaging graph （CLI-graph） to depict the topology characteristics of botnet, which allows SDN controller to locate attacks separately and mitigate the burden of network devices. The theoretical and experimental resuits prove that our scheme is capable of timely botnet detecting in SDNs with the accuracy higher than 90% and the delay less than 56 ms.展开更多
The global growth of the Internet and the rapid expansion of social networks such as Facebook make multilingual sentiment analysis of social media content very necessary. This paper performs the first sentiment analys...The global growth of the Internet and the rapid expansion of social networks such as Facebook make multilingual sentiment analysis of social media content very necessary. This paper performs the first sentiment analysis on code-mixed Bambara-French Facebook comments. We develop four Long Short-term Memory（LSTM）-based models and two Convolutional Neural Network（CNN）-based models, and use these six models, Na？ve Bayes, and Support Vector Machines（SVM） to conduct experiments on a constituted dataset. Social media text written in Bambara is scarce. To mitigate this weakness, this paper uses dictionaries of character and word indexes to produce character and word embedding in place of pre-trained word vectors. We investigate the effect of comment length on the models and perform a comparison among them. The best performing model is a one-layer CNN deep learning model with an accuracy of 83.23 %.展开更多
Wearable devices usually work together with smart phones.To ensure only legitimate smart phones can read the data,they must conduct pairing to establish a shared key.Traditional pairing methods require that the pairin...Wearable devices usually work together with smart phones.To ensure only legitimate smart phones can read the data,they must conduct pairing to establish a shared key.Traditional pairing methods require that the pairing devices have a keyboard or screen for user interaction.However,due to the size limitation,keyboards or screens are hard to be installed in the wearable devices.To solve this problem,we propose a novel pairing method by using ambient sound and light.In this new scheme,any pairing request from smart phone will trigger wearable device vibration.Only after users press the confirm key on the device can the pairing process continues.Then pairing devices collect ambient sound and light at the predetermined time and establish a shared key by using the Diffie-Hellman protocol.To protect against potential man-in-the-middle attacks in the key establishment process,an improved interlock protocol with sound and light comparison is conducted to authenticate the key.If both the sound and light collected by the pairing devices are similar enough,the key is accepted.Otherwise,it is rejected.Compared with current context based pairing methods,our scheme does not impose strict synchronization on devices to collect ambient context data.Moreover,our scheme need not collect and exchange contextual information for multiple times to resist offline brute force attacks.The experimental results and security analysis prove the effectiveness of our scheme.展开更多
基金Supported by the National Natural Science Foundation of China(61272451,61572380)
文摘The distributed detection of botnets may induce heavy computation and communication costs to network devices. Each device in related scheme only has a regional view of Internet, so it is hard to detect botnet comprehensively. In this paper, we propose a lightweight real-time botnet detection framework called Bot-Guard, which uses the global landscape and flexible configurability of software defined network （SDN） to identify botnets promptly. SDN, as a new network framework, can make centralized control in botnet detection, but there are still some challenges in such detections. We give a convex lens imaging graph （CLI-graph） to depict the topology characteristics of botnet, which allows SDN controller to locate attacks separately and mitigate the burden of network devices. The theoretical and experimental resuits prove that our scheme is capable of timely botnet detecting in SDNs with the accuracy higher than 90% and the delay less than 56 ms.
基金Supported by the National Natural Science Foundation of China(61272451,61572380,61772383 and 61702379)the Major State Basic Research Development Program of China(2014CB340600)
文摘The global growth of the Internet and the rapid expansion of social networks such as Facebook make multilingual sentiment analysis of social media content very necessary. This paper performs the first sentiment analysis on code-mixed Bambara-French Facebook comments. We develop four Long Short-term Memory（LSTM）-based models and two Convolutional Neural Network（CNN）-based models, and use these six models, Na？ve Bayes, and Support Vector Machines（SVM） to conduct experiments on a constituted dataset. Social media text written in Bambara is scarce. To mitigate this weakness, this paper uses dictionaries of character and word indexes to produce character and word embedding in place of pre-trained word vectors. We investigate the effect of comment length on the models and perform a comparison among them. The best performing model is a one-layer CNN deep learning model with an accuracy of 83.23 %.
基金Supported by the National Natural Science Foundation of China(61272451,61572380)
文摘Wearable devices usually work together with smart phones.To ensure only legitimate smart phones can read the data,they must conduct pairing to establish a shared key.Traditional pairing methods require that the pairing devices have a keyboard or screen for user interaction.However,due to the size limitation,keyboards or screens are hard to be installed in the wearable devices.To solve this problem,we propose a novel pairing method by using ambient sound and light.In this new scheme,any pairing request from smart phone will trigger wearable device vibration.Only after users press the confirm key on the device can the pairing process continues.Then pairing devices collect ambient sound and light at the predetermined time and establish a shared key by using the Diffie-Hellman protocol.To protect against potential man-in-the-middle attacks in the key establishment process,an improved interlock protocol with sound and light comparison is conducted to authenticate the key.If both the sound and light collected by the pairing devices are similar enough,the key is accepted.Otherwise,it is rejected.Compared with current context based pairing methods,our scheme does not impose strict synchronization on devices to collect ambient context data.Moreover,our scheme need not collect and exchange contextual information for multiple times to resist offline brute force attacks.The experimental results and security analysis prove the effectiveness of our scheme.
