Using homomorphic encryption for privacy-preserving clustering of intrusion detection alerts

Reversible Data Hiding of Digital Image Based on Pixel Combination Algorithm

“…Reversible data hiding in encrypted image (RDH-EI) technology is a technology that embeds secret information in encrypted images in a reversible way, and after the information is extracted losslessly, the original image can be restored losslessly [15], which is the combination and development of reversible information hiding technology and encryption technology. Based on the original intention of privacy protection, the RDH-EI algorithm applied to cloud storage scenarios should provide users with a simple and secure image encryption method on the user side to protect user privacy; in order to facilitate the management of ciphertext images by cloud service providers, the RDH-EI algorithm should be provided with large hidden capacity; at the receiving end, high-quality decrypted images and restored original images that are completely consistent with the original images are reasonable requirements that the RDH-EI algorithm should meet [16]. Most of the existing encrypted image reversible information hiding algorithms are based on the symmetric encryption domain, use stream cipher to encrypt the original image bitwise XOR, and use the spatial correlation of the image to reconstruct the original image.…”

Network Intrusion Detection Based on an Improved Long-Short-Term Memory Model in Combination with Multiple Spatiotemporal Structures

“…When the characteristics of each node not only correspond to different variables but also contain spatiotemporal information, the convolution operation is performed on the basis of the spatiotemporal relationship in the Internet, and then, the extracted features can have both the temporal and spatial characteristics of the process. In the process of training, the network can coordinate the learning of temporal and spatial relationships to obtain the intrusion characteristics of spatiotemporal information in the fusion process [29]. Therefore, the soft measurement application can be realized on the basis of collaborative learning of process spatiotemporal characteristics.…”

“…In the training process of the deep network, the LSTM and the spatiotemporal model can cooperate with each other in the learning of time series characteristics and spatial characteristics [29] and have the advantages of both. First, the process variables are input into p separate channels, and each channel represents a variable, and the LSTM layer is used to extract the timing features for different variables; then, the timing features extracted by each channel are used as nodes 4…”

“…The interface between spatiotemporal and LSTM neural networks requires that the input of the LSTM neural network is a sequence of feature vectors. In other words, it is composed of a collection of feature vectors at consecutive T time steps [29]. Therefore, the input sequence of LSTM must be constructed before training.…”

Efficient and Privacy-Preserving Collaborative Intrusion Detection Using Additive Secret Sharing and Differential Privacy

“…Existing solutions for privacy-preserving collaborative intrusion detection use homomorphic encryption [2], [3] or secure multi-party computation [4], [5] techniques. The idea is to use these techniques to correlate or aggregate alerts generated from local IDS.…”