Search

Many organizations devote significant resources to building high-fidelity deep learning (DL) models. Therefore, they have a great interest in making sure the models they have trained are not appropriated by others. Embedding watermarks (WMs) in DL models is a useful means to protect the intellectual property (IP) of their owners. In this paper, we propose KeyNet, a novel watermarking framework that satisfies the main requirements for an effective and robust watermarking. In KeyNet, any sample in a WM carrier set can take more than one label based on where the owner signs it. The signature is the hashed value of the owner&rsquo, s information and her model. We leverage multi-task learning (MTL) to learn the original classification task and the watermarking task together. Another model (called the private model) is added to the original one, so that it acts as a private key. The two models are trained together to embed the WM while preserving the accuracy of the original task. To extract a WM from a marked model, we pass the predictions of the marked model on a signed sample to the private model. Then, the private model can provide the position of the signature. We perform an extensive evaluation of KeyNet&rsquo, s performance on the CIFAR10 and FMNIST5 data sets and prove its effectiveness and robustness. Empirical results show that KeyNet preserves the utility of the original task and embeds a robust WM.

Outsourcing computation allows wielding huge computational power. Even though cloud computing is the most usual type of outsourcing, resorting to idle edge devices for decentralized computation is an increasingly attractive alternative. We tackle the problem of making peer honesty and thus computation correctness self-enforcing in decentralized computing with untrusted peers. To do so, we leverage the co-utility property, which characterizes a situation in which honest co-operation is the best rational option to take even for purely selfish agents; in particular, if a protocol is co-utile, it is self-enforcing. Reputation is a powerful incentive that can make a P2P protocol co-utile. We present a co-utile P2P decentralized computing protocol that builds on a decentralized reputation calculation, which is itself co-utile and therefore self-enforcing. In this protocol, peers are given a computational task including code and data and they are incentivized to compute it correctly. Based also on co-utile reputation, we then present a protocol for federated learning, whereby peers compute on their local private data and have no incentive to randomly attack or poison the model. Our experiments show the viability of our co-utile approach to obtain correct results in both decentralized computation and federated learning.

The application of algal–bacterial symbiotic interaction as a low-cost and environment-friendly alternative for wastewater treatment was investigated in this research. In this context, a photobioreactor working under sequencing batch mode (SBR) was used. It consists of three basins of 200 L which were fed with pre-treated domestic wastewater. Following the adaptation period, and according to the obtained results, a synergistic interaction between algal and bacterial cultures was developed, which ultimately lead to enhance the removal of different pollutants. During daylight cycles, the dissolved oxygen concentration was around 6 mg/L, which promoted aerobic microbial activities to degrade organic pollutants such that more than 60% of the COD was removed. Additionally, more than 40% and 90% of the nitrogen (NH 4 + -N) and phosphorus were removed, respectively. Based on multivariate data analysis, it was noticed that there is a high positive correlation between DO and NO 3 − (i.e. 0.97), such that, any increase in the DO was associated with an increase in NO 3 − , whereas, a high negative correlation (i.e. -.092) between DO and COD was observed. Generally, the obtained results hold great promise for future process development regarding pollutants removal and the positive contribution in reducing greenhouse gas emissions as reactor mechanical aeration was totally replaced by algal photosynthesis.

It is necessary for the oil and gas industry to drill deeper due to decrease of easily accessible natural reserves. Temperatures of deep wells can exceed 210°C, and conventional cooling and heat extraction techniques are impractical in such a harsh environment. Reliable electronic designs that can sustain high temperature become necessary. This article presents a high-temperature passive radio frequency (RF) mixer for downhole communications. The proposed mixer is designed to upconvert or downconvert the incoming signal with low conversion loss (CL), high linearity, and reliable operation at the ambient temperature up to 250°C. GaN is a wide-bandgap technology that can provide a reliable operation at high ambient temperatures, and the proposed mixer adopts a commercial GaN high-electron-mobility transistor. Measurement results indicate that the proposed mixer achieves a CL of 7.1 dB at local oscillator (LO) power of 2.5 dBm for the downconversion from 230–253 to 97.5 MHz at 250°C and the input P1dB compression point lies at 5 dBm. The designed mixer also achieves 24.5 dB RF-to-intermediate frequency (IF) isolation and 28 dB LO-to-IF isolation at 250°C. The power dissipation of the mixer is virtually zero.

BexZnyO thin films were deposited on crystalline Si (100) and glass substrates at temperature T < 52 °C using RF magnetron sputtering. The X-ray diffraction result indicates that all films have an amorphous nature. By using X-ray photoelectron spectroscopy, the concentrations of Be and Zn atoms were determined. The optical conductivity (σopt) and the dielectric constants of the amorphous BexZnyO (a–BexZnyO) films were determined in the wavelength range 138–1650 nm. The electrical properties of the films for different Be concentrations were also investigated. Changes in the measured of the optical conductivity, the dielectric constants, and the electrical properties values by changing the Be doping level are described for a-BexZnyO films. The adding of Be content to the a-ZnO films has a strong effect on their optical and electrical properties which can be modified by controlling the amount of Be as a dopant. Thus, the a-BeZnO films are good candidates for achieving σopt modulation to values larger than a-ZnO films (1.53 × 1014 1/s, at 3.35 eV). Also, the σopt values were found very small above the bandgap energy, indicating that the films are transparent. The films showed an insulating nature with a huge resistivity of 7.05 × 104 Ω cm for a-Be0·13Zn0.38O0.52 film while the resistivity of a-ZnO film (without doping) exhibits the lowest value of 0.35 Ω cm. Furthermore, the correlation between electrical and optical conductivities of a-BexZnyO films for different Be concentrations has been studied. The amorphous films revealed high σele and low σopt when the Be concentration is < 13%.

As the circuit complexity increases, the number of internal nodes increases proportionally, and individual internal nodes are less accessible due to the limited number of available I/O pins. To address the problem, we proposed power line communications (PLCs) at the IC level, specifically the dual use of power pins and power distribution networks for application/ observation of test data as well as delivery of power. A PLC receiver presented in this paper intends to demonstrate the proof of concept, specifically the transmission of data through power lines. The main design objective of the proposed PLC receiver is the robust operation under variations and droops of the supply voltage rather than high data speed. The PLC receiver is designed and fabricated in CMOS 0.18- $\mu \text{m}$ technology under a supply voltage of 1.8 V. The measurement results show that the receiver can tolerate a voltage drop of up to 0.423 V for a data rate of 10 Mb/s. The power dissipation of the receiver is 3.26 mW under 1.8 V supply, and the core area of the receiver is 74.9 $\mu {\rm m} \times 72.2~\mu \text{m}$ .

This paper presents a high temperature voltage-controlled oscillator (VCO) for downhole communications. The proposed VCO adopts the common-source Colpitts oscillator configuration. The VCO is prototyped using 0.25 μm GaN RF transistors due to the high junction temperature and high unity gain frequency. Two GaN varactors are used to achieve the tuning frequency range of 40 MHz up to 230 oC. The measurements at 230 oC show that the VCO has an output power of 17.5 dBm with ±0.5 dB variations over the tuning range from 320.8 to 360.2 MHz. The measured phase noise is −121 dBc/Hz at 100 kHz offset from 343.3 MHz carrier. The maximum power consumption of the VCO is 122.5 mW.

The decay of easily accessible reserves pushes the oil and gas industry to explore deeper wells, where the ambient temperature often exceeds 210 °C. The need for high temperature operation, combined with real-time data logging, results in a growing demand for robust, high temperature RF circuits. This paper presents a high temperature IF variable gain amplifier (VGA) for downhole communications, which is capable of operating at 230 °C. The proposed VGA is prototyped using GaN on SiC HEMT technology. Measured results at 230 °C show that the VGA has a peak gain of 27 dB at the center frequency of 97.5 MHz, and a dynamic range of 29.4 dB. The input P1dB compression at the peak gain is −11.57 dBm at 230 °C, and −3.63 dBm at 25 °C. The input and output return losses are above 12 dB across the entire temperature range from 25 °C to 230 °C. The maximum power gain and dynamic range drop by 1 dB and 4.7 dB, respectively, at 230 °C. The maximum power dissipation of the VGA is 176 mW under the peak gain at 230 °C.

As oil and gas industry drills deeper, temperatures of wells can exceed 210 °C. Conventional cooling and heat extraction techniques are impractical in the harsh environment, which necessitates reliable electronic systems to operate at high temperatures. This paper presents a high temperature RF downconversion mixer targeted for downhole communications. The proposed mixer is developed with a commercial Gallium Nitride (GaN) high electron mobility transistor (HEMT). The RF band for the mixer is 230–253 MHz and the IF frequency is 97.5 MHz. The mixer achieves the conversion gain of 6 dB at LO power of −10 dBm at temperature 250 °C, and the input P1dB compression point is −7 dBm. It dissipates 50 mW under 5 V supply. The proposed mixer achieves a relatively high conversion gain and good linearity at temperature of 250 °C.

IntroductionEmpathy is critical to the development of professionalism in medical students, but evidence suggests that empathy actually declines over the course of undergraduate medical education.ObjectivesImproving medical student empathy by encouraging students to think about the person behind the illness.MethodsTwo interventions were studied. From December 2015 until November 2016, a fourth year psychiatry medical student book club was conducted. Students were asked to read an autobiography of a lived experience of psychosis. The old age simulation suit aims to simulate the sensory and physical impairments faced by older adults with age related illnesses. A training session provided a transient experience of old age for the students.ResultsForty-four students completed the feedback on the book club. Twenty-eight (64%) stated that they strongly agreed with the statement ‘the book club encouraged me to consider the person behind the illness’. Thirty-nine (89%) stated that after attending the book club their empathy towards people with mental health problems had increased. Eleven students completed full feedback following the old age simulation session. Empathy statements relating to living in an ageing body improved from the pre-test median score of 4 (range 1–7) to a median score of 6 (range 2–8) post-teaching session. Empathy statements focusing on sensory and physical impairments had pre-test score median of 3 (range 1–7) and post-test median 8 (range 3–9).ConclusionsFeedback from these sessions has demonstrated that with a little creativity, empathy training can be delivered to medical students with a positive impact.Disclosure of interestThe authors have not supplied their declaration of competing interest.

Magnetoelectric (ME) multiferroic bismuth manganite (BiMnO[Formula: see text] has ferroelectric (FE) and ferromagnetic order parameters that coupled biquadratically. Landau free energy expression in these order parameters is proposed to investigate the phase transition and thermodynamic properties of a BiMnO3 film. The surface energy term is included in the free energy. We find that the ME coupling enhances the magnetic transition temperature. The magnetization and polarization order parameters are both increased strongly below the new magnetic transition temperature. The curves for magnetization and polarization versus temperature show that the phase transition is second-order. The entropy, free energy, and specific heat of the BiMnO3 multiferroic are calculated. Thickness-driven phase transitions for magnetization and polarization of BiMnO3 thin film are clearly indicated.

Federated learning (FL) is an alternative to centralized machine learning (ML) that builds a model across multiple decentralized edge devices (a.k.a. workers) that own the training data. This has two advantages: i) the data used for training are not uploaded to the server and ii) the server can distribute the training load across the workers instead of using its own resources. However, due to the distributed nature of FL, the server has no control over the behaviors of the workers. Malicious workers can, therefore, orchestrate different kinds of attacks against FL. Byzantine attacks are amongst the most common and straightforward attacks on FL. They try to prevent FL models from converging by uploading random updates. Several techniques have been proposed to detect such kind of attacks, but they usually entail a high cost for the server. This hampers one of the main benefits of FL, which is load reduction. In this work, we propose a highly efficient approach to detect workers that try to perform Byzantine FL attacks. In particular, we analyze the last layer biases of deep learning (DL) models on the server side to detect malicious workers. We evaluate our approach with two deep learning models on the MNIST and CIFAR-10 data sets. Experimental results show that our approach significantly outperforms current methods in runtime while providing similar attack detection accuracy.