Real Power Traces Research Articles

Deep Learning-based Attacks on Masked AES Implementation

<p>To ensure the confidentiality of the message, the AES (Advanced Encryption Standard) block cipher algorithm can be widely used. Furthermore, an implementation of masked AES is often used to resist side-channel attacks. To recover secret keys embedded in cryptographic devices with masked AES, we present some side-channel attacks based on deep learning models in profiling and non-profiling scenarios. The proposed method which applies the mask value profiling technique represents new approaches for extracting the secret key. To defeat the masked AES implementation, deep learning models such as multi-layer perceptron and convolutional neural networks are developed. In a non-profiling scenario, we adopt the DDLA (Differential Deep Learning Analysis) to extract sensitive information such as the secret key. The main idea of our method is that it is possible to adopt a new binary labeling method to conduct the DDLA based on the HW (Hamming Weight) model. We show several experiments using real power traces measured from the ChipWhisperer platform in profiling attacks and the ASCAD dataset in non-profiling attacks respectively. Whether we target naïve or masked AES implementation, the experimental results show the predominant key recovery accuracy.</p> <p> </p>

SmoothOperator

With the ever growing popularity of cloud computing and web services, Internet companies are in need of increased computing capacity to serve the demand. However, power has become a major limiting factor prohibiting the growth in industry: it is often the case that no more servers can be added to datacenters without surpassing the capacity of the existing power infrastructure. In this work, we first investigate the power utilization in Facebook datacenters. We observe that the combination of provisioning for peak power usage, highly fluctuating traffic, and multi-level power delivery infrastructure leads to significant power budget fragmentation problem and inefficiently low power utilization. To address this issue, our insight is that heterogeneity of power consumption patterns among different services provides opportunities to re-shape the power profile of each power node by re-distributing services. By grouping services with asynchronous peak times under the same power node, we can reduce the peak power of each node and thus creating more power head-rooms to allow more servers hosted, achieving higher throughput. Based on this insight, we develop a workload-aware service placement framework to systematically spread the service instances with synchronous power patterns evenly under the power supply tree, greatly reducing the peak power draw at power nodes. We then leverage dynamic power profile reshaping to maximally utilize the headroom unlocked by our placement framework. Our experiments based on real production workload and power traces show that we are able to host up to 13% more machines in production, without changing the underlying power infrastructure. Utilizing the unleashed power headroom with dynamic reshaping, we achieve up to an estimated total of 15% and 11% throughput improvement for latency-critical service and batch service respectively at the same time, with up to 44% of energy slack reduction.

전력 피크 감소를 위한 스케줄링 알고리즘의 성능 평가 및 분석

Peak power reduction is becoming increasingly important not only for grid operators but also for residential users. The scheduling of electric loads tries to reduce the power peak by splitting the power-on period of an electric device into multiple smaller ones and by interleaving the on-periods of every device in a holistic way. This paper analyzes the performance of EDF, LSF, TCBM, and lazy scheduling algorithms and proposes the enhancement schemes. For analysis, we have implemented the scheduling policies in a simulation environment for distributed control systems. Through extensive experiments using real power traces, we discuss their performance characteristics in terms of power deviation, switch count, and temperature violation ratio. To prevent excessive switching, we propose to employ the concept of limited preemptibility and evaluate its effect on performance. It is found that the best performance is achieved when the scheduler capacity is dynamically adjusted to the actual power demand. The experiment results show that, by load scheduling, the probability of having a power deviation greater than 150W is reduced from 21.5% down to 3.2%.

EnergyBox: Disclosing the wireless transmission energy cost for mobile devices

While evolving mobile technologies bring millions of users closer to the vision of information anywhere-anytime, device battery depletions still hamper the quality of experience to a great extent. The energy consumption of data transmission is highly dependent on the traffic pattern, and we argue that designing energy efficient data transmissions starts by energy awareness. Our work proposes EnergyBox, a parametrised tool that facilitates accurate and repeatable energy consumption studies for 3G and WiFi transmissions at the user end using real traffic data.The tool takes as input the parameters of a network operator and the power draw for a given mobile device in the 3G and WiFi transmission states. It outputs an estimate of the consumed energy for a given packet trace, either synthetic or captured in a device using real applications. Using nine different applications with different data patterns the versatility and accuracy of the tool was evaluated. The evaluation was carried out for a modern and popular smartphone in the WiFi setting, a specific mobile broadband module for the 3G setting, and within the operating environment of a major mobile operator in Sweden. A comparison with real power traces indicates that EnergyBox is a valuable tool for repeatable and convenient studies. It exhibits an accuracy of 94–99% for 3G, and 95–99% for WiFi given the studied applications’ traces.Next the tool was deployed in a use case where a location sharing application was ran on top of two alternative application layer protocols (HTTP and MQTT) and with two different data exchange formats (JSON and Base64). The illustrative use case helped to identify the appropriateness of the pull and push strategies in sharing location data, and the benefit of EnergyBox in characterising where the breaking point lies for preferring one or the other protocol, under which network load, or exchange data format.