Our world is more connected than ever. A user of technology products produces data with every interaction, transaction, and movement. Our digitized world affords endless opportunities to discover new insights, correlations, and discoveries through data science and artificial intelligence (A.I.).
However, along with those benefits also come some serious risks. The power and control of data custodians is being scrutinized, and whether limits should be placed on how data is used.
In response to these concerns, governments have begun to regulate data collection, use, and privacy. It is completely reasonable to set common standards related to data control and use, since the consequences of data being in the wrong hands is potentially enormous, like the consequence to an unwitting consumer from identity theft.
However, taking regulations too far can stifle innovation. Developers of technologies, like artificial intelligence, can utilize specific mathematical techniques in algorithms to alleviate these security and privacy concerns.
A.I. is grounded in mathematics, and, simply described, is a string of algorithms chained seamlessly together. Differential privacy and cryptographic techniques can be used within the chain to provide information anonymity and a secure method of transfer.
Cryptography is the study of techniques used for secure communication in the presence of adversarial actors. It uses a set of techniques to encrypt communication using complex algorithms to convert the information into a unique code.
Differential privacy is like cryptography in that it uses complex mathematical techniques to mask information so that it may be transferred, stored, or used in a secure manner.
Differential privacy uses a set of algorithmic rules that limit the disclosure of specific information within a database, or more complex techniques creating mathematical ‘noise’ or randomization within the algorithm. The addition of ‘noise’ or randomization to the output of an algorithm makes it difficult to reverse engineer and determine specific information from a data-set.
Both, differential privacy and asymmetric cryptography require simply adding another algorithm within an A.I. model.
Blockchain would be another effective method to ensure security and privacy, but its complexity would not make it a good fit for all applications. It is also a much more complex method and would be costly to maintain. Like asymmetric cryptography, it also converts information into cryptographic code. It requires a network of computers to validate the computations of the other users within the network, making this technique costly over time from the requisite computing power necessary to perform the validation. Depending on the use and sensitivity of the information being used, blockchain might be a better solution because with its complexity comes additional security.
In sum, the scope of data regulations need not be cumbersome in order to achieve the desired level of security and privacy. The use of mathematical techniques like differential privacy, asymmetric cryptography, and blockchain will resolve many of the concerns these regulations attempt to address.