SEARCH · Search NASA

Results for “poisoning attack”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

3 records

All sources NASA NTRS DOE OSTI

NASA NTRSDate not supplied

NNREPAIR: Constraint-based Repair of Neural Network Classifiers

We present NNREPAIR, a constraint-based technique for repairing neural network classifiers by i) fixing the logic of the network at an intermediate layer, and ii) fixing the decision conditions of the network at the last layer. NNREPAIR first uses fault localization to find potentially faulty network parameters (specifically weights) and then performs repair using constraint solving to apply small modifications to remedy the defects. We present novel strategies to enable precise yet efficient repair such as inferring correctness specifications to act as oracles for intermediate layer repair, and generation of experts wrt each class. We demonstrate the technique in the context of three different scenarios: (1) Improving the over all accuracy of a model, (2) Fixing security vulnerabilities caused by poisoning of training data and (3) Improving the robustness of the network against adversarial attacks. Our evaluation on MNIST and CIFAR-10 models shows that NNREPAIR can improve the accuracy by 45.56% on poisoned data and 11.11% on adversarial data. NNREPAIR also provides improvement, albeit modest, in the overall accuracy of models, without requiring new data or re-training.

Repair↗

NASA NTRS1959

Design and Operating Criteria for Fluorine Disposal by Reaction with Charcoal

Experiments with the charcoal-fluorine reaction for the disposal of fluorine have shown generally that this method is effective over a wide range of conditions. Pure fluorine or fluorine diluted with nitrogen to concentrations as low as 0.3 percent fluorine may be disposed of efficiently within the rate limitation. Maximum feed rates have been established and are inversely proportional to the charcoal-bed particle diameter. Moisture content in the charcoal had no appreciable effect on the disposal efficiency after the reaction zone was established and the moisture was driven off by the heat of reaction. There was no evidence of bed poisoning resulting from continued use. Design parameters may be based on the stoichiometric requirements plus sufficient excess charcoal to maintain desired efficiency toward the end of a disposal operation. The length of time a given reactor may be used continuously is limited by the rate of fluorine input and the resistance of the system to heat and fluorine attack. Refractory-lined reactors have been in routine field use at the Lewis Research Center for over a year and have given satisfactory service over a wide range of conditions.

Schmidt, Harold W.↗

NASA NTRS2011

Two Catalysts for Selective Oxidation of Contaminant Gases

Two catalysts for the selective oxidation of trace amounts of contaminant gases in air have been developed for use aboard the International Space Station. These catalysts might also be useful for reducing concentrations of fumes in terrestrial industrial facilities especially facilities that use halocarbons as solvents, refrigerant liquids, and foaming agents, as well as facilities that generate or utilize ammonia. The first catalyst is of the supported-precious-metal type. This catalyst is highly active for the oxidation of halocarbons, hydrocarbons, and oxygenates at low concentrations in air. This catalyst is more active for the oxidation of hydrocarbons and halocarbons than are competing catalysts developed in recent years. This catalyst completely converts these airborne contaminant gases to carbon dioxide, water, and mineral acids that can be easily removed from the air, and does not make any chlorine gas in the process. The catalyst is thermally stable and is not poisoned by chlorine or fluorine atoms produced on its surface during the destruction of a halocarbon. In addition, the catalyst can selectively oxidize ammonia to nitrogen at a temperature between 200 and 260 C, without making nitrogen oxides, which are toxic. The temperature of 260 C is higher than the operational temperature of any other precious-metal catalyst that can selectively oxidize ammonia. The purpose of the platinum in this catalyst is to oxidize hydrocarbons and to ensure that the oxidation of halocarbons goes to completion. However, the platinum exhibits little or no activity for initiating the destruction of halocarbons. Instead, the attack on the halocarbons is initiated by the support. The support also provides a high surface area for exposure of the platinum. Moreover, the support resists deactivation or destruction by halogens released during the destruction of halocarbons. The second catalyst is of the supported- metal-oxide type. This catalyst can selectively oxidize ammonia to nitrogen at temperatures up to 400 C, without producing nitrogen oxides. This catalyst converts ammonia completely to nitrogen, even when the concentration of ammonia is very low. No other catalyst is known to oxidize ammonia selectively at such a high temperature and low concentration. Both the metal oxide and the support contribute to the activity and selectivity of this catalyst.

Wright, John D.↗