A security manager is deciding whether to implement a new machine learning technique for fraud detection at a financial institution. Based on historical data, for every 1,000 transactions, 12 of them will be fraudulent. However, all techniques sometimes misidentify a legitimate transaction as fraudulent and raise an alert. After running a test of the machine learning technique. the manager estimates that for every 1,000 transactions the detection rate would be as follows: Attempted Fraudulent Transaction -- 3 12 --... 9?4 17" Total 26 1 ,000 According to the data, if a fraud alert is raised for a transaction, what is the probability that the transaction is legitimate? Round to the nearest thousandth. A. 0.017 B. 0.654 C. 0.346 D. 0.529