Nowadays, smart devices such as sensors and actuators have been extensively deployed in the smart home system. These devices have communication capability, which lets the house user remotely control automated electronic devices by entering a single command or a PIN number [1]. To save time or energy, sometimes, house users prefer to hire third parties, such as home security companies that provide 24/7 security protection services. However, the participation of the sensing application and the security company could potentially leak the privacy of the house users. Actually, without well-designed privacy preservation solutions, it is possible that private data is misused.

The privacy leakage for smart home [2]—even for very simple appliances such as electronic locks (e-Lock) [3]—is a potential challenge. It increasingly affects all house users, given the fact that captured private data can be misused to infer personal activities. The insight is based on the observation that some intermittent activities, such as e-Lock switched on/off, could possibly infer personal absence/presence at the smart home.

A false alarm attack system in [3] is studied, which is launched against the most popular and commercially endorsed electronic lock. However, in the context of smart home security monitoring system, privacy preservation and privacy analyses (e.g., [4]) for e-Lock have not been presented. More importantly, it lacks a formal model of privacy analysis—privacy guarantee and quantitative evaluation are desirable. The pertinent privacy-related questions should be addressed regarding time-series e-Lock state data:

1) How much privacy is lost, and to what extent if all e-Lock states in smart home are open for access?

2) If perturbation methods which introduce uncertain noise to true personal data are deployed, could aggregator still query the leakage of privacy?

3) If the aggregator (e.g., server or the cloud of the security company) is not trusted, could we protect the privacy by an access control scheme based on information flow?

Our contributions: we propose a theoretical privacy framework to analyze the privacy leakage through accommodating fundamental functionality, e.g., sharing/hiding, perturbation and access control for aggregated time-series e-Lock state dataset for smart home system stored in cloud computing. We not only carefully study potential privacy inference but also try to address corresponding concerns about privacy loss in the case of the true dataset, of the distorted dataset and of the dataset under information flow protection:

1) To offer privacy protection, noisy perturbing of real data is assumed. We then invoke the differential privacy method to analyze the significant difference between specific residences’ absences from the smart home.

2) When the switching on/off operations of e-Locks are open to access, they are treated as the input and, in turn, could be modeled as a real-valued correlated Gaussian random variable. Based on that, a hidden Markov Chain model is provided to measure the absence of occupancies in correlation with the e-Lock’s switching on/off operations.

3) Turning on/off operations of e-Locks are transmitted to the security companies’ servers or even their cloud. They may be borrowed by a third party for investigation in future. It is possible that the aggregator is untrusted. An attempt to utilize cyber property information flow is taken to shield privacy.

Currently, there are different ways of integrating smart homes into the broader context of smart services, smart grids, and even smart cities, considering that our world is growing smarter than ever.

This ongoing trend has come up with interesting and useful applications. However, in the light of a majority of consumers being lack of technology and not necessarily trained in security in computer system, cloud computing system, or control systems, this paper has extensively studied the security of privacy needs. When certain sensory data is leaked or the communication data pattern is revealed in the current smart home system, consumers or network managers should be notified or be aware of the privacy leakage.

Meanwhile, due to its limited computing capacity, a single smart home cannot process all data in time. Therefore, taking advantage of cloud computing becomes a reasonable choice. However, privacy leakage that occurs during the data transmission between the smart home system and the cloud environment becomes a big concern. Due to this potential issue, a framework is proposed in this paper to integrate smart homes into the platform clouds or service clouds. In the proposed framework, to simply use the privacy protection solution, we only think about the e-lock, which is relatively simple. Although other smart devices, e.g., refrigerator, oven, etc., are more complicated devices, they can be protected in a similar way.

E-Locks in a smart home include three components—keypad, central processing unit and solenoid (actuator). If the credential inputted at keypad is valid, a signal is sent to solenoid to change the e-Lock’s state from off to on. The application generates time-series categorical data, which is aggregated to the server/cloud of the security company through secure communication channel. The admin or even authority staff (e.g., policeman) may be able to query or even access the anonymized dataset from now on.

Note that the framework proposed in this paper only focuses on theoretical analyses by using some privacy preservation methodologies. How to implement them in real-world smart home system (e.g., e-lock) is out of the scope of this paper.

Let us take an electronic lock (e-lock) as an example: at the time point t i , a few controllers, O h , O p , and O c representing the resident, physical automation control device in the smart room and the cyber remote control program, respectively, issue their own control command { c O h , c O p , c O c } t i . In the proposed framework, since each of O h , O p , and O c could be malicious, we will choose the right control command O r for the e-lock P based on the current context set φ ( t i ) and the e-lock statuses, S k = [ M ] i × j . The context set φ ( t i ) could include different kinds of residents such as host, guests, and so on. Note that the privacy-related issues can include many more factors and therefore could be more complicated than the scenario we are discussing in this paper. For example, the security company may also hire different kinds of security guards, full-time, contract, substitution, etc. They all have been granted different levels of access control privilege based on their priority. This paper only thinks about a very simple case study and will count them in for our future research. In other words, the context, together with the current status of the e-lock, decides whether privacy is leaked or not. In future, our work may also include simulation or even real-world experiments. Comparing with other privacy-related research, this paper provides more theoretical analyses. Furthermore, our works will centrally focus on privacy preservation via perturbing distributed noisy information to time-series e-Lock state change data to minimize the privacy loss with a lower utility-privacy tradeoff. In addition, how to extend the hidden Markov Chain method to precisely quantify privacy loss and corresponding counter-measures via differentially private protection will be studied.