Resilient PHEV Charging Policies under Price Information Attacks Yifan Li † , Ran Wang † , Ping Wang † , Dusit Niyato † , Walid Saad ‡ , and Zhu Han * † School of Computer Engineering, Nanyang Technological University, Singapore ‡ ECE Department, University of Miami, Coral Gables, FL, USA ∗ ECE Department, University of Houston, Houston, TX, USA Abstract—Enabling a bidirectional energy flow between power grids and plug-in hybrid electric vehicles (PHEVs) using vehicle- to-grid (V2G) and grid-to-vehicle (G2V) communications is considered as one of the key components of the future smart grid. On the one hand, the PHEV owner needs to charge its PHEV through the grid, given possibly time-varying electricity pricing schemes. On the other hand, the energy stored in a PHEV can also be sold back to the grid so as to act as an ancillary service while possibly generating revenues to its owner. Consequently, this motivates the need to develop smart charging policies that enable the PHEV owner to optimally decide on when to charge or discharge its vehicle, while minimizing its long-term energy consumption cost. In this paper, we model this PHEV energy management problem as a Markov decision process (MDP), which is solved by using a linear programming (LP) technique so as to obtain the optimal charging policy. In particular, we devise optimal charging policies that are resilient to the price information attacks such as denial of service (DoS) attacks and price manipulation attacks over the grid’s communication network. We show that, under potential price information attacks, each PHEV can optimize its charging policies given only an estimated price information, which leads to a discrepancy between the real and expected costs. To this end, we analyze this cost difference using the proposed MDP model, which can also guide the system designer and administrator to decide whether reinforcing the system’s security is required. The simulation results show that the proposed PHEV charging policy is effective and is adaptable to different PHEV mobility patterns, battery levels and varying electricity prices. It is also demonstrated that improving the system’s ability to detect and resolve the attack can obviously reduce the impact brought by the attacks. I. I NTRODUCTION Nowadays, with the growing concern in reducing reliance on fossil fuel to decrease the greenhouse gas emissions and relieve the problem of global climate change, reelectrification of the automobile transportation is attracting more and more attention from both academia and industry, resulting in a fast development of plug-in hybrid electric vehicles (PHEVs) [1], [2]. The possibility of a two-way communication between PHEVs and the power grid enables the PHEVs to play a dual role in the electricity market. Whenever the PHEVs need to connect to the grid for charging purposes, they are seen as energy consumers. In contrast, whenever the PHEVs are used as ancillary services that can feed energy back to the grid, they can be seen as energy providers. Hence, the PHEV owner needs to optimally decide when to act as energy consumers and as energy providers in order to minimize the long-term energy consumption cost. Thus, a well-designed charging policy enabling the PHEV owner to strategically charge or discharge the PHEV is important. Taking into account all practical factors such as the PHEV mobility, variation of the battery level and time-varying electricity prices, designing smart and resilient charging policies is challenging [3]. Recently, several models for PHEV energy management have been proposed in the literature. In [4], a real-time V2G control algorithm for parked vehicles under price uncertainty is proposed, which is used to adapt the control operation to the hourly notified price information, aiming at maximizing the profit for the owner of the vehicle during the entire parking time. In [5], a daily energy cost minimization problem for the vehicle owners is considered. A dynamic programming model is applied to formulate and solve the problem, and a state- dependent double-threshold policy is proposed and proved to be optimal. Other PHEV charging works are found in [3], [6]. However, little has been done to study charging policies while taking potential price information attacks into con- sideration, when designing PHEV energy management poli- cies. For instance, although a bidirectional communications infrastructure can bring many benefits to the smart grid, it can introduce new vulnerabilities. For example, a malicious attacker can attempt to tap into the grid’s communication system with the aim to cause malfunctions to the power grid, disrupt the electricity market, or make monetary profits [7], [8]. One easy target for such attacks is the manipulation of the real-time pricing information that is communicated by the public utility to the vehicles [8]. The attacker may disrupt the transmission of the electricity price information to the PHEV owner, resulting in the loss of the pricing information, which is, in fact, one of the possible denial-of-service (DoS) attacks on the smart grid. Alternatively, it is possible for the attacker to manipulate the pricing information by injecting incorrect price values so as to compromise the charging policies of the PHEV owners. The main contribution of this paper is that, to study the energy consumption cost minimization problem for a PHEV owner under potential price information attack, and analyze the impact of such an attack (e.g., how much cost the PHEV owner will suffer from the attack), we formulate the problem as a Markov decision process (MDP) and we devise optimal and resilient PHEV charging policies. We show that the derived result is helpful for the system operator or administrator to decide whether to reinforce the security system to effectively guard the PHEV against such attacks. Our simulation results show that the proposed PHEV charging policy is effective and is adaptable to different PHEV mobility patterns, battery levels and varying electricity prices. It is also verified that improving the system’s ability to detect and IEEE SmartGridComm 2012 Symposium - Cyber Security and Privacy 978-1-4673-0911-0/12/$31.00 ©2012 IEEE 389