The network intrusion detection system is an important tool for protecting computer networks against threats and malicious attacks. Many techniques have recently been proposed; however, these face significant challenges due to the continuous emergence of new threats that are not recognized by existing systems. In this paper, we propose a novel two-stage deep learning (TSDL) model, based on a stacked auto-encoder with a soft-max classifier, for efficient network intrusion detection. The model comprises two decision stages: an initial stage responsible for classifying network traffic as normal or abnormal, using a probability score value. This is then used in the final decision stage as an additional feature, for detecting the normal state and other classes of attacks. The proposed model is able to learn useful feature representations from large amounts of unlabeled data and classifies them automatically and efficiently. To evaluate its effectiveness, several experiments are conducted on two public datasets, specifically the benchmark KDD99 and UNSW-NB15 datasets. Comparative simulation results demonstrate that our proposed model significantly outperforms existing approaches, achieving high recognition rates, up to 99.996% and 89.134%, for the KDD99 and UNSW-NB15 datasets respectively. We conclude that our model has the potential to serve as a future benchmark for the deep learning and network security research communities. INDEX TERMS Computational intelligence, two-stage deep learning model, feature representation, network intrusion detection, stacked auto-encoder.

Portocaval hemitransposition maybe useful in liver transplantation when hepatopetal flow to the liver graft cannot be established by other techniques. Rescue after failure of conventional technique was not possible in two patients.

Insulin is known to regulate pancreatic ␤-cell function through the activation of cell surface insulin receptors, phosphorylation of insulin receptor substrate (IRS)-1 and -2, and activation of phosphatidylinositol (PI) 3-kinase. However, an acute effect of insulin in modulating ␤-cell electrical activity and its underlying ionic currents has not been reported. Using the perforated patch clamp technique, we found that insulin (1-600 nmol/l) but not IGF-1 (100 nmol/l) reversibly hyperpolarized single mouse ␤-cells and inhibited their electrical activity. The dose-response relationship for insulin yielded a maximal change (mean ؎ SE) in membrane potential of ؊13.6 ؎ 2.0 mV (P < 0.001) and a 50% effective dose of 25.9 ؎ 0.1 nmol/l (n ‫؍‬ 63). Exposing patched ␤-cells within intact islets to 200 nmol/l insulin produced similar results, hyperpolarizing islets from ؊47.7 ؎ 3.3 to ؊65.6 ؎ 3.7 mV (P < 0.0001, n ‫؍‬ 11). In single cells, insulin-induced hyperpolarization was associated with a threefold increase in whole-cell conductance from 0.6 ؎ 0.1 to 1.7 ؎ 0.2 nS (P < 0.001, n ‫؍‬ 10) and a shift in the current reversal potential from ؊25.7 ؎ 2.5 to ؊63.7 ؎ 1.0 mV (P < 0.001 vs. control, n ‫؍‬ 9; calculated K ؉ equilibrium potential ‫؍‬ ؊90 mV). The effects of insulin were reversed by tolbutamide, which decreased cell conductance to 0.5 ؎ 0.1 nS and shifted the current reversal potential to ؊25.2 ؎ 2.3 mV. Insulin-induced ␤-cell hyperpolarization was sufficient to abolish intracellular calcium concentration ([Ca 2؉ ] i ) oscillations measured in pancreatic islets exposed to 10 mmol/l glucose. The application of 100 nmol/l wortmannin to inactivate PI 3-kinase, a key enzyme in insulin signaling, was found to reverse the effects of 100 nmol/l insulin. In cell-attached patches, single ATP-sensitive K ؉ (K ATP ) channels were activated by bath-applied insulin and subsequently inhibited by wortmannin. Our data thus demonstrate that insulin activates the K ATP channels of single mouse pancreatic ␤-cells and islets, resulting in membrane hyperpolarization, an inhibition of electrical activity, and the abolition of [Ca 2؉ ] i oscillations. We thus propose that locally released insulin might serve as a negative feedback signal within the islet under physiological conditions.

A novel calcium-dependent potassium current (Kslow) that slowly activates in response to a simulated islet burst was identified recently in mouse pancreatic β-cells (Göpel, S.O., T. Kanno, S. Barg, L. Eliasson, J. Galvanovskis, E. Renström, and P. Rorsman. 1999. J. Gen. Physiol. 114:759–769). Kslow activation may help terminate the cyclic bursts of Ca2+-dependent action potentials that drive Ca2+ influx and insulin secretion in β-cells. Here, we report that when [Ca2+]i handling was disrupted by blocking Ca2+ uptake into the ER with two separate agents reported to block the sarco/endoplasmic calcium ATPase (SERCA), thapsigargin (1–5 μM) or insulin (200 nM), Kslow was transiently potentiated and then inhibited. Kslow amplitude could also be inhibited by increasing extracellular glucose concentration from 5 to 10 mM. The biphasic modulation of Kslow by SERCA blockers could not be explained by a minimal mathematical model in which [Ca2+]i is divided between two compartments, the cytosol and the ER, and Kslow activation mirrors changes in cytosolic calcium induced by the burst protocol. However, the experimental findings were reproduced by a model in which Kslow activation is mediated by a localized pool of [Ca2+] in a subspace located between the ER and the plasma membrane. In this model, the subspace [Ca2+] follows changes in cytosolic [Ca2+] but with a gradient that reflects Ca2+ efflux from the ER. Slow modulation of this gradient as the ER empties and fills may enhance the role of Kslow and [Ca2+] handling in influencing β-cell electrical activity and insulin secretion.

The results suggest that dose and timing of DBMC infusions may be important variables affecting allograft survival. A randomized prospective trial is now in progress to compare group 3 DBMC infusion protocol with controls receiving OLTX alone.