Full-privacy secured search engine empowered by efficient genome-mapping algorithms

Since the 90s, keyword-based search engines have been helping people locate relevant web content via a simple query, so have the recent full-text-based search engines mainly used for plagiarism detection following an article upload. However, these "free" or paid services operate by storing users' search queries and preferences for personal profiling and targeted ads delivery, while user-uploaded articles can further profit the service providers as part of their expanding databases. In short, search engine privacy has not been an option for web exploration in the past decades. Here we demonstrate that a database or internet search, provided with the entire article as a query, can be correctly carried out without revealing users' sensitive queries by an irreversible encoding scheme and an efficient FM-index search routine that is generally used in the NGS of genomes. In our solution, Sapiens Aperio Veritas Engine (S.A.V.E.), every word in the query is encoded into one of 12 "amino acids" (a.a.) comprising a pseudo-biological sequence (PBS) at users' local machines. The PBS-mediated plagiarism detection is done by users' submission of locally encoded PBS through our cloud service to locate identical duplicates in the collected web contents which had been encoded in the same way as the query. It is found that PBSs with a length longer than 12 a.a., can return correct results with a false positive rate <0.8%. S.A.V.E. runs at a similar speed as Bowtie and is 4 orders faster than BLAST. S.A.V.E., functioning in both regular and in-private search modes, provides a new option for efficient internet search and plagiarism detection in a compressed search space without a chance of storing and revealing users' confidential contents. We expect that future privacy-aware search engines can reference the ideas proposed herein. S.A.V.E. is made available at https://dyn.life.nthu.edu.tw/SAVE/

A Leaf Recognition Algorithm for Plant Classification Using Probabilistic Neural Network

In this paper, we employ Probabilistic Neural Network (PNN) with image and data processing techniques to implement a general purpose automated leaf recognition algorithm. 12 leaf features are extracted and orthogonalized into 5 principal variables which consist the input vector of the PNN. The PNN is trained by 1800 leaves to classify 32 kinds of plants with an accuracy greater than 90%. Compared with other approaches, our algorithm is an accurate artificial intelligence approach which is fast in execution and easy in implementation.