Deep Learning Classification Methods Applied to Tabular Cybersecurity Benchmarks

This research recasts the network attack dataset from UNSW-NB15 as an intrusion detection problem in image space. Using one-hot-encodings, the resulting grayscale thumbnails provide a quarter-million examples for deep learning algorithms. Applying the MobileNetV2’s convolutional neural network architecture, the work demonstrates a 97% accuracy in distinguishing normal and attack traffic. Further class refinements to 9 individual attack families (exploits, worms, shellcodes) show an overall 54% accuracy. Using feature importance rank, a random forest solution on subsets shows the most important source-destination factors and the least important ones as mainly obscure protocols. It further extends the image classification problem to other cybersecurity benchmarks such as malware signatures extracted from binary headers, with an 80% overall accuracy to detect computer viruses as portable executable files (headers only). Both novel image datasets are available to the research community on Kaggle.

Image Classifiers for Network Intrusions

This research recasts the network attack dataset from UNSW-NB15 as an intrusion detection problem in image space. Using one-hot-encodings, the resulting grayscale thumbnails provide a quarter-million examples for deep learning algorithms. Applying the MobileNetV2’s convolutional neural network architecture, the work demonstrates a 97% accuracy in distinguishing normal and attack traffic. Further class refinements to 9 individual attack families (exploits, worms, shellcodes) show an overall 56% accuracy. Using feature importance rank, a random forest solution on subsets show the most important sourcedestination factors and the least important ones as mainly obscure protocols. The dataset is available on Kaggle.

Game-Centred Study Using Eigenvector Centrality in High-Level Women’s Volleyball: Play Efficacy is Independent of Game Patterns… Or is it?

In sports, it is often assumed that distinct game patterns may influence the outcome of the play differently. However, a few articles about men’s volleyball have suggested that play efficacy may rely more on the quality of individual attack actions, and not on game patterns. Therefore, the goal of this paper was to scrutinize if and how game patterns influence play efficacy in high-level women’s volleyball. Eigenvector Centrality was assessed to integrate direct and indirect relationships between games actions. Thirteen matches from the women’s World Grand Prix’2015 were analysed (46 sets; 2,016 plays). Actions were categorized according to game complex (K0 to KV) and three levels of the efficacy of each play: error, continuity, and point. The results showed that play efficacy was independent of game patterns (the central pattern was non-ideal setting conditions in all complexes and preference for using slow attacks in the extremities of the net). There were, however, some regularities for each game complex. For example, while in KI to KIII, Zone 4 was the most used attack zone, in KIV and KV there was a complete inversion to Zone 2. Moreover, results revealed that women’s volleyball games are more predictable in relation to the play space (attack zones) while increasing the risk through enhanced game speed (attack tempo), in comparison with what studies in men’s volleyball have shown. Future studies should consider situational variables (e.g., match status, home vs away matches), and individual players’ actions should be considered in order to understand their relationships with team patterns better.

Evaluation of computer network security using attack undirected geography

<span>To secure a wealth of data traversing the computer network at your fingertips is compulsory. But when attack arises at various parts of the network it is difficult to protect, especially when each incident is investigated separately. Geography is a necessary construct in computer networks. The analytics of geography algorithms and metrics to curate insight from a security problem are a critical method of analysis for computer systems. A geography based representation is employed to highlight aspects (on a local and global level) of a security problem which are Eigenvalue, eccentricity, clustering coefficient and cliques. Network security model based on attack undirected geography (AUG) is familiarized. First, analysis based upon association rules is presented then the attack threshold value is set from AUG. The probability of an individual attack edge and associated network nodes are computed in order to quantify the security threat. The simulation is exploited to validate that results are effective.</span>

Graph state-based quantum authentication scheme

Inspired by the special properties of the graph state, a quantum authentication scheme is proposed in this paper, which is implemented with the utilization of the graph state. Two entities, a reliable party, Trent, as a verifier and Alice as prover are included. Trent is responsible for registering Alice in the beginning and confirming Alice in the end. The proposed scheme is simple in structure and convenient to realize in the realistic physical system due to the use of the graph state in a one-way quantum channel. In addition, the security of the scheme is extensively analyzed and accordingly can resist the general individual attack strategies.

CRYPTANALYSIS OF THE QSDC PROTOCOL WITHOUT USING PERFECT QUANTUM CHANNEL

We study the security of quantum secure direct communication without perfect quantum channel. Our analysis is focused on a general individual attack, i.e. entangle-ancilla attack. The sufficient and necessary condition for a successful eavesdropping operation is obtained, and a particular attack is presented, by which Eve can elicit the whole secret without being discovered. Finally we find the root of this loophole and give a possible improvement of this protocol accordingly.

A TWO-STEP CHANNEL-ENCRYPTING QUANTUM KEY DISTRIBUTION PROTOCOL

A two-step channel-encrypting quantum key distribution protocol is proposed. Using the previously shared EPR pairs as the quantum key, two bits of classical key can be established via one information carrier EPR state on average. In theory, the efficiency of this protocol reaches 100%, and there is no need to consume any entangled states including both the quantum key and the information carriers in ideal condition. The protocol can resist the particular attack that is fatal to other some channel-encrypting schemes. Principally, we prove the security against the most general individual attack of this protocol. Entanglement collapse in practical situation, as well as the realistic implementation of this protocol is also discussed.