Real-Time Anomaly Detection in Edge Streams

Given a stream of graph edges from a dynamic graph, how can we assign anomaly scores to edges in an online manner, for the purpose of detecting unusual behavior, using constant time and memory? Existing approaches aim to detect individually surprising edges. In this work, we propose Midas , which focuses on detecting microcluster anomalies , or suddenly arriving groups of suspiciously similar edges, such as lockstep behavior, including denial of service attacks in network traffic data. We further propose Midas -F, to solve the problem by which anomalies are incorporated into the algorithm’s internal states, creating a “poisoning” effect that can allow future anomalies to slip through undetected. Midas -F introduces two modifications: (1) we modify the anomaly scoring function, aiming to reduce the “poisoning” effect of newly arriving edges; (2) we introduce a conditional merge step, which updates the algorithm’s data structures after each time tick, but only if the anomaly score is below a threshold value, also to reduce the “poisoning” effect. Experiments show that Midas -F has significantly higher accuracy than Midas . In general, the algorithms proposed in this work have the following properties: (a) they detects microcluster anomalies while providing theoretical guarantees about the false positive probability; (b) they are online, thus processing each edge in constant time and constant memory, and also processes the data orders-of-magnitude faster than state-of-the-art approaches; and (c) they provides up to 62% higher area under the receiver operating characteristic curve than state-of-the-art approaches.

Raw Biogas as Feedstock for the OCM Process

The paper presents the research results obtained in the process of oxidative coupling of methane, in which unpurified biogas was used as the feedstock. Biogas obtained from two kinds of biomass materials, i.e., plant materials (potato and beet pulp, Corn-Cob-Mix—biogas 1) and animal waste (waste from fish filleting—biogas 2) was considered. The influence of temperature, the ratio of methane/oxygen and total flows of feedstock on the catalytic performance in oxidative coupling of methane process was investigated. Comparative tests were carried out using pure methane and a mixture of methane-carbon dioxide to simulate the composition of biogas 2. The process was carried out in the presence of an Mn-Na2WO4/SiO2 catalyst. Fresh and used catalysts were characterised by means of powder X-ray diffraction, X-ray photoelectron spectroscopy, and low-temperature nitrogen adsorption techniques. In oxidative coupling of methane, the type of raw material used as the source of methane has a small effect on methane conversion (the differences in methane conversion are below 3%), but a significant effect on the selectivity to C2. Depending on the type of raw material, the differences in selectivity to C2 reach as high as 9%. However, the Mn-Na2WO4/SiO2 catalyst operated steadily in the tested period of time at any feedstock composition. Moreover, it was found that CO2, which is the second main component of biogas in addition to methane, has an effect on catalytic performance. Comparative results of catalytic tests indicate that the CO2 effect varies with temperature. Below 1073 K, CO2 exerts a small poisoning effect on methane conversion, while above this temperature the negative effect of CO2 disappears. In the case of selectivity to C2+, the negative effect of CO2 was observed only at 1023 K. At higher temperatures, CO2 enhances selectivity to C2+. The effect of CO2 was established by correlating the catalytic results with the temperature programmed desorption of CO2 investigation. The poisoning effect of CO2 was connected with the formation of surface Na2CO3, whose concentration depends on temperature.

Algorithms for Detecting and Preventing Attacks on Machine Learning Models in Cyber-Security Problems

Machine learning algorithms can be vulnerable to many forms of attacks aimed at leading the machine learning systems to make deliberate errors. The article provides an overview of attack technologies on the models and training datasets for the purpose of destructive (poisoning) effect. Experiments have been carried out to implement the existing attacks on various models. A comparative analysis of cyber-resistance of various models, most frequently used in operating systems, to destructive information actions has been prepared. The stability of various models most often used in applied problems to destructive information influences is investigated. The stability of the models is shown in case of poisoning up to 50% of the training data.

Poisoning Effect of CO: How It Changes Hydrogen Electrode Reaction and How to Analyze It Using Differential Polarization Curve

The hydrogen electrode reaction (HER) on Pt electrode in a H2SO4 solution when CO gas was injected/stopped was studied using polarization resistance curve [...]

Stroboscopic operando spectroscopy of the dynamics in heterogeneous catalysis by event-averaging

Heterogeneous catalyst surfaces are dynamic entities that respond rapidly to changes in their local gas environment, and the dynamics of the response is a decisive factor for the catalysts’ action and activity. Few probes are able to map catalyst structure and local gas environment simultaneously under reaction conditions at the timescales of the dynamic changes. Here we use the CO oxidation reaction and a Pd(100) model catalyst to demonstrate how such studies can be performed by time-resolved ambient pressure photoelectron spectroscopy. Central elements of the method are cyclic gas pulsing and software-based event-averaging by image recognition of spectral features. A key finding is that at 3.2 mbar total pressure a metallic, predominantly CO-covered metallic surface turns highly active for a few seconds once the O2:CO ratio becomes high enough to lift the CO poisoning effect before mass transport limitations triggers formation of a √5 oxide.

The poisoning effect of K, Ca and Na on CeZrTiAl catalyst with nanosheet skeleton structure

The poisoning effect of KNO3, NaNO3, and Ca(NO3)2 on CeZrTiAl catalyst for selective catalytic reduction of NO with NH3 was investigated. It was found that the activity deactivation rate follows K> Na > Ca. SEM and BET showed that the accumulation of catalysts was severe after poisoning, and the nanosheet γ-Al2O3 skeleton structure disappeared due to alkali coating. The decrease of the specific surface area is accompanied by pore blockage, making the catalyst unable to expose rich reaction sites. In addition, the fewer surface Ce3+ and chemisorbed oxygen on the surface of the poisoned catalyst weaken the cycle between Ce3+ and Ce4+, resulting in bad redox performance. Thus, the failure to realize the efficient oxidation of NO to NO2. Another critical reason for catalyst poisoning failure is that the decrease of surface acid sites seriously affects the adsorption and activation of NH3 and NOx on the catalyst surface.