The parallel learning in neural networks can greatly shorten the training time. Its prior efforts were mostly limited to distributing inputs to multiple computing engines. It is because the gradient descent algorithm in the neural network training is inherently sequential. This paper proposes a novel CNN parallel training method for image recognition. It overcomes the sequential property of the gradient descent and enables the parallel training with the speculative backpropagation. We found that the Softmax and ReLU outcomes in the forward propagation for the same labels are likely to be very similar. This characteristic makes it possible to perform the forward and backward propagation simultaneously. We implemented the proposed parallel model with CNNs in both software and hardware, and evaluated its performance. The parallel training reduces the training time by 34% in CIFAR-100 without the loss of the prediction accuracy compared to the sequential training. In many cases, it even improves the accuracy.

Infrared (IR) communication is one of the wireless communication methods mainly used to manipulate consumer electronics devices. Traditional IR devices support only simple operations such as changing TV channels. These days, consumer electronic devices such as smart TV are connected to the internet with the introduction of IoT. Thus, the user’s sensitive information such as credit card number and/or personal information could be entered with the IR remote. This situation raises a new problem. Since TV and the set-top box are visual media, these devices can be used to control and/or monitor other IoT devices at home. Therefore, personal information can be exposed to eavesdroppers. In this paper, we experimented with the IR devices’ reception sensitivity using remotes. These experiments were performed to measure the IR reception sensitivity in terms of distance and position between the device and the remote. According to our experiments, the transmission distance of the IR remote signal is more than 20 m. The experiments also revealed that curtains do not block infrared rays. Consequently, eavesdropping is possible to steal the user’s sensitive information. This paper proposes a simple, practical, and cost-effective countermeasure against eavesdropping, which does not impose any burden on users. Basically, encryption is used to prevent the eavesdropping. The encryption key is created by recycling a timer inside the microcontroller typically integrated in a remote. The key is regenerated whenever the power button on a remote is pressed, providing the limited lifecycle of the key. The evaluation indicates that the XOR-based encryption is practical and effective in terms of the processing time and cost.

Blockchain is a promising solution for Industry 4.0; however, it does not guarantee input data integrity. We propose a field-programmable gate array (FPGA)-based private blockchain system for the industrial Internet of Things, where the transaction generation is performed inside the FPGA in an isolated and enclaved manner.

This paper designed and implemented a low-cost surveillance and information system for museums with two new coding schemes using Visible Light Communication (VLC). For security, we use a key in place of the clock in the Manchester coding. For the enhanced data transfer, we use the unbalanced time duration for data one and zero. The prototype system was built with two low-cost microcontrollers, an LED, and linear light sensors. With 1 m distance in a dark environment, the maximum data transfer rate was roughly 1.93 kbits/s with a 10 Φ white LED. Under the dark-to-modest ambient brightness of up to 120 lux, the white LEDs were performed well in terms of the transmission distance, compared to LEDs with different colors. The experiments with flashlights reveal that our system can easily detect the prohibited use of flashlight. Experiments with the mobile phones’ flashlight also revealed that photographing with the flashlight would have a worse impact on arts than the continuous illumination. We also measured the distance between arts to avoid inevitable interference between illumination sources in neighbor.

In cloud computing, users can rent computing resources from service providers according to their demand. Spot instances are unreliable resources provided by cloud computing services at low monetary cost. When users perform tasks on spot instances, there is an inevitable risk of failures that causes the delay of task execution time, resulting in a serious deterioration of quality of service (QoS). To deal with the problem on spot instances, we propose an estimated interval-based checkpointing (EIC) using weighted moving average. Our scheme sets the thresholds of price and execution time based on history. Whenever the actual price and the execution time cross over the thresholds, the system saves the state of spot instances. The Bollinger Bands is adopted to inform the ranges of estimated cost and execution time for user's discretion. The simulation results reveal that, compared to the HBC and REC, the EIC reduces the number of checkpoints and the rollback time. Consequently, the task execution time is decreased with EIC by HBC and REC. The EIC also provides the benefit of the cost reduction by HBC and REC, on average. We also found that the actual cost and execution time fall within the estimated ranges suggested by the Bollinger Bands.