Many IoT systems generate a huge and varied amount of data that need to be processed and responded to in a very short time. One of the major challenges is the high energy consumption due to the transmission of data to the cloud. Edge computing allows the workload to be offloaded from the cloud at a location closer to the source of data that need to be processed while saving time, improving privacy, and reducing network traffic. In this paper, we propose an energy efficient approach for IoT data collection and analysis. First of all, we apply a fast error-bounded lossy compressor on the collected data prior to transmission, that is considered to be the greatest consumer of energy in an IoT device. In a second phase, we rebuild the transmitted data on an edge node and process it using supervised deep learning techniques. To validate our approach, we consider the context of driving behavior monitoring in intelligent vehicle systems where vital signs data are collected from the driver using a Wireless Body Sensor Network (WBSN) and wearable devices and sent to an edge node for stress level detection. The experimentation results show that the amount of transmitted data has been reduced by up to 103 times without affecting the quality of medical data and driver stress level prediction accuracy.

No abstract

With the exponential growth in Internet-of-Things (IoT) devices, security and privacy issues have emerged as critical challenges that can potentially compromise their successful deployment in many data-sensitive applications. Hence, there is a pressing need to address these challenges, given that IoT systems suffer from different limitations, and IoT devices are constrained in terms of energy and computational power, which renders them extremely vulnerable to attacks. Traditional cryptographic algorithms use a static structure that requires several rounds of computations, which leads to significant overhead in terms of execution time and computational resources. Moreover, the problem is compounded when dealing with multimedia contents, since the associated algorithms have stringent QoS requirements. In this paper, we propose a lightweight cipher algorithm based on a dynamic structure with a single round that consists of simple operations, and that targets multimedia IoT. In this algorithm, a dynamic key is generated and then used to build two robust substitution tables, a dynamic permutation table, and two pseudo-random matrices. This dynamic cipher structure minimizes the number of rounds to a single one, while maintaining a high level of randomness and security. Moreover, the proposed cipher scheme is flexible as the dimensions of the input matrix can be selected to match the devices' memory capacity. Extensive security tests demonstrated the robustness of the cipher against various kinds of attacks. The speed, simplicity and high-security level, in addition to low error propagation, make of this approach a good encryption candidate for multimedia IoT devices.

The protection of multimedia content has become a key area of research, since very often a user's privacy and confidentiality can be at risk. Although a large number of image encryption algorithms have recently emerged, only a subset of these algorithms are suitable for real applications. These algorithms however use non-integer operations such as chaotic solutions that introduce a sizeable overhead in terms of latency and resources, in addition to floating-point hardware that is costly to implement. Designing an efficient, lightweight, and secure image encryption algorithm is still a hard challenge; yet, it is crucial to have in order to meet the demands of recent multimedia applications running on energy-limited devices. In this paper, an efficient image encryption scheme based on a dynamic structure is proposed. The structure of the proposed cipher consists of two different lightweight rounds (forward and backward chaining blocks) and a block permutation process. In addition, a key derivation function is proposed to produce a dynamic key based on a secret key and a nonce. This key, according to its configuration, can be changed for each validate time (session) or for each new input image. Then, based on this key, the cipher layers are produced, which are an integer or a binary diffusion matrix and a substitution table S-box, together with a permutation table P-box. The proposed dynamic cipher is designed to provide high robustness against contemporary powerful attacks, and permits reducing the required number of rounds for achieving the lightweight property. Experimental simulations demonstrate the efficiency and robustness levels of the proposed scheme.

No abstract

Underwater wireless sensor networks (UWSNs) have recently been proposed as a way to observe and explore aquatic environments. Sensors in such networks are used to perform pollution monitoring, disaster prevention or assisted navigation and to send monitored data to the sink. Compared to traditional sensor networks, sensors in UWSNs consume more energy due to the acoustic technology used in under water communications. Node clustering is a common method to organize data traffic and reduce in-network communications while improving scalability and energy consumption. In this paper, we present a new clustering method to handle the spatial similarity between node readings. We suppose that readings are sent periodically from sensor nodes to their appropriate cluster-heads (CHs). Then a two tier data aggregation technique is proposed. At the first level, each node periodically cleans its readings in order to eliminate redundancies before sending its data set to its CH. Once the CH receives all data sets, it applies an enhanced K-means algorithm based on a one-way ANOVA model to identify nodes generating identical data sets and to aggregate these sets before sending them to the sink. Our proposed approach is validated via experiments on real sensor data and comparison with other existing clustering and data aggregation techniques.

No abstract

Reservoir Computing is an attractive paradigm of recurrent neural network architecture, due to the ease of training and existing neuromorphic implementations. Successively applied on speech recognition and time series forecasting, few works have so far studied the behavior of such networks on computer vision tasks. Therefore we decided to investigate the ability of Echo State Networks to classify the digits of the MNIST database. We show that even if ESNs are not able to outperform state-of-the-art convolutional networks, they allow low error thanks to a suitable preprocessing of images. The best performance is obtained with a large reservoir of 4,000 neurons, but committees of smaller reservoirs are also appealing and might be further investigated.