arXiv subjects

Vladimir Dyo

Publications and source records attributed to Vladimir Dyo.

5 recordsLinked to original sources

arXiv2022

Behaviour-neutral Smart Charging of Plugin Electric Vehicles: Reinforcement learning approach

High-powered electric vehicle (EV) charging can significantly increase charging costs due to peak-demand charges. This paper proposes a novel charging algorithm which exploits typically long plugin sessions for domestic chargers and reduces the overall charging power by boost charging the EV for a short duration, followed by low-power charging for the rest of the plugin session. The optimal parameters for boost and low-power charging phases are obtained using reinforcement learning by training on EV's past charging sessions. Compared to some prior work, the proposed algorithm does not attempt to predict the plugin session duration, which can be difficult to accurately predict in practice due to the nature of human behavior, as shown in the analysis. Instead, the charging parameters are controlled directly and are adapted transparently to the user's charging behavior over time. The performance evaluation on a UK dataset of 3.1 million charging sessions from 22,731 domestic charge stations, demonstrates that the proposed algorithm results in 31% of aggregate peak reduction. The experiments also demonstrate the impact of history size on learning behavior and conclude with a case study by applying the algorithm to a specific charge point.

eess.SY↗

arXiv2021

Privacy-preserving Identity Broadcast for Contact Tracing Applications

Wireless Contact tracing has emerged as an important tool for managing the COVID19 pandemic and relies on continuous broadcasting of a person's presence using Bluetooth Low Energy beacons. The limitation of current contact tracing systems in that a reception of a single beacon is sufficient to reveal the user identity, potentially exposing users to malicious trackers installed along the roads, passageways, and other infrastructure. In this paper, we propose a method based on Shamir secret sharing algorithm, which lets mobile nodes reveal their identity only after a certain predefined contact duration, remaining invisible to trackers with short or fleeting encounters. Through data-driven evaluation, using a dataset containing 18 million BLE sightings, we show that the method drastically reduces the privacy exposure of users. Finally, we implemented the approach on Android phones to demonstrate its feasibility and measure performance for various network densities.

cs.CR↗

arXiv2020

Battery-assisted Electric Vehicle Charging: Data Driven Performance Analysis

As the number of electric vehicles rapidly increases, their peak demand on the grid becomes one of the major challenges. A battery-assisted charging concept has emerged recently, which allows to accumulate energy during off-peak hours and in-between charging sessions to boost-charge the vehicle at a higher rate than available from the grid. While prior research focused on the design and implementation aspects of battery-assisted charging, its impact at large geographical scales remains largely unexplored. In this paper we analyse to which extent the battery-assisted charging can replace high-speed chargers using a dataset of over 3 million EV charging sessions in both domestic and public setting in the UK. We first develop a discrete-event EV charge model that takes into account battery capacity, grid supply capacity and power output among other parameters. We then run simulations to evaluate the battery-assisted charging performance in terms of delivered energy, charging time and parity with conventional high-speed chargers. The results indicate that in domestic settings battery-assisted charging provides 98% performance parity of high-speed chargers from a standard 3 kW grid connection with a single battery pack. For non-domestic settings, the battery-assisted chargers can provide 92% and 99% performance parity of high-speed chargers with 10 battery packs using 3kW and 7kW grid supply respectively.

eess.SY↗

arXiv2020

Practical Hash-based Anonymity for MAC Addresses

Given that a MAC address can uniquely identify a person or a vehicle, continuous tracking over a large geographical scale has raised serious privacy concerns amongst governments and the general public. Prior work has demonstrated that simple hash-based approaches to anonymization can be easily inverted due to the small search space of MAC addresses. In particular, it is possible to represent the entire allocated MAC address space in 39 bits and that frequency-based attacks allow for 50% of MAC addresses to be enumerated in 31 bits. We present a practical approach to MAC address anonymization using both computationally expensive hash functions and truncating the resulting hashes to allow for k-anonymity. We provide an expression for computing the percentage of expected collisions, demonstrating that for digests of 24 bits it is possible to store up to 168,617 MAC addresses with the rate of collisions less than 1%. We experimentally demonstrate that a rate of collision of 1% or less can be achieved by storing data sets of 100 MAC addresses in 13 bits, 1,000 MAC addresses in 17 bits and 10,000 MAC addresses in 20 bits.

cs.CR↗

arXiv2020

Cross Hashing: Anonymizing encounters in Decentralised Contact Tracing Protocols

During the COVID-19 (SARS-CoV-2) epidemic, Contact Tracing emerged as an essential tool for managing the epidemic. App-based solutions have emerged for Contact Tracing, including a protocol designed by Apple and Google (influenced by an open-source protocol known as DP3T). This protocol contains two well-documented de-anonymisation attacks. Firstly that when someone is marked as having tested positive and their keys are made public, they can be tracked over a large geographic area for 24 hours at a time. Secondly, whilst the app requires a minimum exposure duration to register a contact, there is no cryptographic guarantee for this property. This means an adversary can scan Bluetooth networks and retrospectively find who is infected. We propose a novel "cross hashing" approach to cryptographically guarantee minimum exposure durations. We further mitigate the 24-hour data exposure of infected individuals and reduce computational time for identifying if a user has been exposed using $k$-Anonymous buckets of hashes and Private Set Intersection. We empirically demonstrate that this modified protocol can offer like-for-like efficacy to the existing protocol.

cs.CR↗