SIT719 Taxonomy of Attacks, Defenses and Consequences in Adversarial Machine Learning Assignment Sample

• Subject Code : SIT719
• University : Deakin University My Assignment Services is not sponsored or endorsed by this college or university.
• Subject Name : IT Computer Science

Security and Privacy Issues in Analytics

Quasi-identifiers are a subset of an attribute which themselves are not distinctive, however when correlated with an entity, it happens to re-identify and has greater chance of reducing uncertainty. If each quasi-identifier is combined in right manner it holds the maximum probability of becoming personally identifying information.

Well, it is shown that postal codes, gender and birth dates are not unique identity of an individual. However, if these information are assembled, they are capable enough to predict identity of 87% of individual in US.

Many situations in business framework reels around transferring of data which usually refers to put data base as risk. The best suited practice to tackle the situation came up with the idea of anonymizing the data in 1998 when Pierangela Samarati and Latanya Sweeney induced a scientifically guaranteed data where none of the records where re-identified while the data practically remained useful. The data is said to have k-anonymity property, where the respondents remove or encrypt the attributes making it next to impossible to distinguish from at least k-1 individual in the database provided.

K- anonymity privacy model was introduced to mark the risk of re-identification to an individual when other subsets are assembled. The k-anonymized data is low- risk, privacy preserving modified database sharing only fraction of information to the concerned, having its own limitations and strength.

If a data owner wants a way to transfer a dataset containing highly sensitive information can employ k-anonymity with the method of suppression and generalization to get satisfactorily result against various breaches in concern to data privacy and security.

A dataset with holding vast information in the form of table with organized rows and columns, where rows are set of records and column are attributes associated with each record has to be shared to external sources, in this situation the responder can encrypt or replaced some of the numeric values of attributes let’s say age with sign “*”, this method of anonymization is said to be suppression. Suppression is a technical model to not to let various quasi- identifiers get organized and formulate a personally identifying information besides contributing to tackle against potential privacy breached. Generalization, is another commonly used method predicting the successful story of k-anonymity against security and privacy issues breakage, where a certain group of attributed are compiled together and replaced with broader category. For, example, instead of defining age of each record, the data can be categorizing by making a broader range of age group between 18-20, 20-22 and so on. K-anonymity is a practice of hiding in the crowd where a certain entity is encrypted under well studied way so that even after implication of various quasi-identifies metrics, the true identity of the record remains hidden. K- anonymity directs the limitation of database while keeping it fully functional to extract the important information also possessing effects by certain anonymized data. Such data has capable enough to challenge various security and privacy issues.

List of Data Anonymization Tools

• ARX
• Amnesia
• µ-ARGUS
• sdcMicro
• Anonimatron

References for Security and Privacy Issues in Analytics

M. Morgenstern. Security and Inference in multilevel database and knowledge based systems. Proc. of the ACM SIGMOD Conference, pages 357--373, 1987.

L. Willenborg and T. De Waal. Statistical Disclosure Control in Practice. SpringerVerlag, 1996.

Remember, at the center of any academic work, lies clarity and evidence. Should you need further assistance, do look up to our Computer Science Assignment Help