Protection of data privacy from vulnerability using two-fish technique with Apriori algorithm in data mining

Protection of data privacy from vulnerability using two-fish technique with Apriori algorithm in data mining

연구 분야: Strategies

학회: The Journal of Supercomputing

The confidential data is mainly managed by creating passwords, tokens, and unique identifiers in an authorized manner. These records must be kept in a safe location away from the reach of unauthorized third parties. Both the client and server sides must be encrypted using the two-fish algorithm, which secures the distinction of private data. By gaining access to the user's information, a data miner may be able to steal it. To avoid such situations, both the data miner and the server must be encrypted. Further, the previous techniques faced several shortcomings in case of higher computational overhead, poor resource utilization, prone to single point failure, lower accuracy, noise, poor security, higher distortion, etc. In this study, both the client and server sides are encrypted using a two-fish algorithm to avoid information loss while transferring data to overcome these problems. The way the state-of-art techniques handled the privacy preservation issue often leads to privacy violations. This paper focuses on mining frequent itemsets present in the medical data by also ensuring privacy. Frequent itemset mining mainly aims to extract highly correlated items from the database and to achieve this novel fruitfly whale optimization algorithm (FWOA) combined with the Apriori algorithm. The Apriori heuristic and bio-inspired algorithms are integrated to solve the frequent itemset problem by reducing the low runtime performance when handling large datasets and also offering high-quality solutions. The adaptive k-anonymity approach is used for preserving data privacy by transforming the original data into an encrypted mode and offering privacy to the top-k frequent itemsets mining. The main advantage of the adaptive k-anonymity approach is that the confidential information disclosed by an individual user cannot be identified from at least k − 1 individuals. We ensure that the proposed methodology can offer data privacy in real time by the experiments conducted in a medical dataset. The experimental results obtained highlight the robustness of this scheme.