By using Parallel Privacy-Preserving Computation in the scenarios of RSA, DSS, Bilinear Paring or ElaGamal Cryptosystem methods one can easily protect online financial transaction security.
Biometric + Password + Identity + Smart Card (Three Factor is best technique to protect any account holder from DDoS attack.
Similarly, if ECC technique is used, it is lightweight and provide more security, we can also develop a mechanism using this technique in which a curve is divided into infinity numbers, during computation one random number will chooses, which is also not known to the owner of smart card & that number will be used for a single transaction.
Computational Diffie-Hellman method is also hard for an adversary to extract personal sensitive information from a user, and we can easily protect the movements, social and financial activities easily. Similarly, anonymity and randomness of a random key (chosen) is mandatory for such sensitive activities when designing scheme for it.
If you need assistant, I am ready to transfer my expertise in this case.
I think the great deal will be to create a method of suspicious transactions detection. So the main research here is to create a data mining method and/or analysis method based on machine learning for such detection.