2.  LiveLab CHALLENGES & SOLUTIONS

We first discuss areas of recognizing LiveLab and provide our solutions in an implementation-agnostic way.

2.1  Comfort

One of the primary issues while developing, implementing, and providing such a extensive logger is privacy. To be able to develop a better understanding of  the privacy issues of members, in particular what details they are reluctant to have signed, we performed several discussions. Obviously, we discovered that participants’ biggest concern was regarding their identification. That is, members do not want scientists to be able to affiliate their details with their details. Amazingly, they are not con-cerned about some potentially delicate details being gathered as long as the details is not straight connected to their identification. Our members were fine with a 1-out-of-25 privacy for much of the details we initially considered personal, including GPS place and web accessibility record. In contrast, they are not comfortable with the material of email or information being gathered straight, considering it highly personal. However, they do not mind if this material is examined and cleaned in the product, as specific below. For our XIAOMI MI4 implementation, we mentioned the logger with our members detailed at a official meeting before the phones were distributed.  

With our participants’ issues recognized, we employ the following techniques to protect privacy while maintaining relevant details for research. First, we make use of one-way hashing to protect the originality of a details entry without exposing its material. For example, we hash the XIAOMI MI3 contact figures members contact. With hashing, we can still build contact research without knowing actual contact figures. Second, we perform details removal in the product. For example, we extract emoticons from e-mails and sms information without gathering the raw material. Finally, we structure the research group so that the details research and logger group does not interact with or even know the members, to prevent connecting details to customer names. A individual human factors group functions as the user interface with our members but does not deal straight with the logger or accessibility the raw details. This enables us to contact the members in a privacy delicate way, which we have discovered to be necessary on numerous occasions.  For example, it has allowed us to gain specific understanding in to utilization styles, allowing us to schedule unplanned discussions with users who display a extreme change in actions.  

2.2  Energy

Collecting details from XIAOMI MI4 mobile phones in the field normally happens upon power expense and decreases battery power life-time. Significantly reduced battery power life-time is likely to effect utilization, thus the utilization details would not perfectly indicate customer actions in the actual world [10, 18]. Therefore, an precise logger must carefully minimize the ability effect of details selection. Towards this goal, we have employed the following four signing techniques to decrease power intake.  

The first is to drive signing with interferes with. Any moment a program occasion occurs, such as a program being started out, or WiFi being turned on, the logger grabs the occasion immediately  and records it.  Interrupt-driven signing prevents regular polling and grabs activities in an exact and immediate way.  

The second is piggy-backing. Modern XIAOMI MI3 smart cellphone systems already log a number of products with timestamps, such as contact record, SMSs, e-mails, etc, often for customer convenience. We can save energy by simply gathering them regularly, ideally while the product is connected in and nonproductive. Collection once a day is a good balance between power intake, details loss, and reviews.   

The third is improving the signing period for regularly signed products. This is especially important when the details is powerexpensive to gather, e.g. GPS numbers and data transfer useage dimensions. It is almost impossible to maintain a reasonable battery power life-time if such dimensions are gathered regularly. Our solution is to trade the length of details selection for regularity of selection. That is, by gathering details more time, we can gather the details less regularly. The reasoning is that a many years will allow the details to be gathered at more times of the day and more places. While such details will not be able to catch certain temporary characteristics, e.g. the place track of a customer in a given day, they still maintain key research regarding the customer and network actions. Additionally, regularity can be dynamically changed based on perspective.  For example if the XIAOMI MI4 cellphone is connected in, moving rapidly, or in a new place, details can be signed more regularly.  

The fourth is hitch-hiking. A logger can hitch-hike on to current program wakeups to decrease extra program wakeups for details selection. XIAOMI MI3 Smartphones normally awaken from the low-power method when the customer triggers the product, when the program gets activities that it is waiting for, e.g. an inbound contact or written text, and when the os needs to carry out certain maintenance. By gathering details when the smart cellphone normally awakens, the logger can prevent an extra wakeup.