Online medicine diagnosis based on pathological images has been regarded as a pervasive method due to the advances in electronic healthcare and Internet of Things (IoT), however, it also causes storage and computing stress on the local IoT devices. To solve this problem, a nature way is to outsource images to cloud servers. Unfortunately, a range of security and privacy issues arise while delegating both storage and computing to the untrusted external servers. In this paper, we present a privacy-preserving feature extraction and retrieval scheme over medical images, which allows images storage and processing on two separate cloud servers. We share the images using secret sharing technology and design a set of secure two-party computation protocols between the two cloud servers. Then a privacy-preserving convolutional neural networks (CNN) framework is constructed to achieve feature extraction, classification and retrieval of images in the encrypted domain. We analyse and evaluate our scheme in terms of both security and efficiency. The results indicate that the proposed secure protocols in our scheme can significantly reduce the computation overhead while protecting the privacy of images as well as the data generated during execution on cloud servers and final results. The performance of our scheme in image feature extraction, classification and retrieval is at a similar level comparable to the scheme based on original CNN in plaintext.

Approximate pattern matching (APM) measures whether the Hamming distance between two strings is less than a threshold value. APM has been widely utilized, such as gene matching and facial recognition. Yet, the genetic data are privacy-sensitive, resulting that the owners are unwilling to share the raw data. This inspires us to explore how to securely perform APM. After revisiting threshold private set intersection (TPSI), we first propose and formalize a functionality named Boolean threshold private set intersection (BTPSI). The new proposed BTPSI primitive returns a Boolean value (0 or 1) to the user, rather than the actual elements in TPSI. We then construct a secure protocol for the BTPSI functionality with semihonest security. Besides, we first combine oblivious transfer and BTPSI to achieve the efficient construction of secure approximate pattern matching (SAPM) protocol in a semihonest model. Furthermore, we implement our SAPM protocol to demonstrate its real practicality. The performance result shows that when the text length is 2 20 ${2}^{20}$ and the pattern length is 2 10 ${2}^{10}$ , the total runtime is less than 3 s.

Secure wildcard pattern matching (WPM) allows the pattern holder to obtain the matched positions without revealing pattern and text information about both parties. However, standard secure WPM may have limitations in practical applications, as users may prefer to have access to the actual data of the match in many scenarios. Fortunately, secure wildcard pattern matching with query (SWMQ) extends standard secure WPM by allowing the pattern holder to obtain the matched positions and the actual data, which has important applications in many scenarios, such as electronic healthcare and gene matching. This also motivates the research of SWMQ in this paper. In this study, we focus on the efficient construction of SWMQ in the semihonest adversary setting. First, we propose two new primitives, hereafter referred to as shared wildcard pattern matching (Sh-WPM) and choice-sharing oblivious transfer (CSOT). Furthermore, we propose an SWMQ protocol via Shared WPM and CSOT. In addition, we evaluate the performance of SWMQ. More specifically, the running time in local area network and wide area network settings is less than 0.4 and 2 s, respectively, when the text length is 2 16 ${2}^{16}$ and the pattern length is 2 12 ${2}^{12}$ . In fact, our evaluation results suggest that SWMQ is not only more broadly functional, but also comparable in efficiency to state-of-the-art approaches.