On the Privacy of TraceTogether, the Singaporean COVID-19 Contact Tracing Mobile App and Recommendations for Australia
Authors (in Alphabetic Order):
Differential Privacy and Publicly Known Constraints
Some papers have argued about a weakness in the definition of differential privacy; namely that it does not provide privacy when the dataset has correlated rows or when some public constraints about the dataset are known. Differential privacy is a mathematical notion of privacy for statistical analysis of sensitive datasets, i.e., datasets containing private data of individuals.
One of the claimed benefits of differential privacy over other alternative definitions, e.g., k-anonymity, data de-identification, is that the privacy guarantee holds even under arbitrary background knowledge of an adversary. The aforementioned weaknesses posit that this property of differential privacy is in fact not true in cases where the dataset has correlated rows or when some public constraints about the dataset are known. For instance, when exact query answers had already been released.
Averaging Attacks on TableBuilder
Recently, we published our paper on arXiv, which shows two attacks on an algorithm used by the Australian Bureau of Statistics (ABS). The algorithm allows people query Australian census data through their tool called TableBuilder.
About this blog
My intention is to write slightly non-technical blog posts about interesting topics from privacy, information security and cryptography. Stay tuned.