System and method for intentionally distorting digital media to reduce the accuracy of generative machine learning algorithms

A method comprising:

• monitoring transmissions originating on an internal network and destined for an external network;
• determining that a transmission of the transmissions comprises an original piece of media associated with a subject, the original piece of media comprising a set of data elements, wherein each data element of the set of data elements comprises a value that lies between a minimum value and a maximum value;
• in response to determining that the transmission comprises the original piece of media associated with the subject:
• intercepting the transmission comprising the original piece of media before the transmission reaches the external network;
• generating a modified piece of media by applying a media modification process to the original piece of media, the media modification process comprising:
• selecting a subset of data elements from the set of data elements of the original piece of media;
• and
• for each data element of the subset of data elements, replacing a value of the data element with a new value of a set of new values, wherein:
• the new value lies between the minimum value and the maximum value;
• and
• at least one of the subset of data elements and the set of new values is chosen such that a first value of an accuracy metric calculated for a first generative machine learning algorithm is less than, by a given factor, a second value of the accuracy metric calculated for a second generative machine learning algorithm;
• wherein:
• the first generative machine learning algorithm comprises a generative model trained to generate synthetic representations of the subject based on media associated with the subject and modified according to the media modification process;
• and
• the second generative machine learning algorithm comprises the generative model trained to generate synthetic representations of the subject based on the media associated with the subject, unmodified according to the media modification process;
• and
• replacing the transmission with a new transmission destined for the external network, the new transmission comprising the modified piece of media.

A non-transitory computer-readable medium comprising instructions that are configured, when executed by a hardware processor, to:

• monitor transmissions originating on an internal network and destined for an external network;
• determine that a transmission of the transmissions comprises an original piece of media associated with a subject, the original piece of media comprising a set of data elements, wherein each data element of the set of data elements comprises a value that lies between a minimum value and a maximum value;
• in response to determining that the transmission comprises the original piece of media associated with the subject:
• intercept the transmission comprising the original piece of media before the transmission reaches the external network;
• generate a modified piece of media by applying a media modification process to the original piece of media, the media modification process comprising:
• selecting a subset of data elements from the set of data elements of the original piece of media;
• and
• for each data element of the subset of data elements, replacing a value of the data element with a new value of a set of new values, wherein:
• the new value lies between the minimum value and the maximum value;
• and
• at least one of the subset of data elements and the set of new values is chosen such that a first value of an accuracy metric calculated for a first generative machine learning algorithm is less than, by a given factor, a second value of the accuracy metric calculated for a second generative machine learning algorithm;
• wherein:
• the first generative machine learning algorithm comprises a generative model trained to generate synthetic representations of the subject based on media associated with the subject and modified according to the media modification process;
• and
• the second generative machine learning algorithm comprises the generative model trained to generate synthetic representations of the subject based on the media associated with the subject, unmodified according to the media modification process;
• and
• replace the transmission with a new transmission destined for the external network, the new transmission comprising the modified piece of media.