More on Computing’s Divided Future

Thank you so much for addressing the danger posed by the current Chinese government's imperialistic ambitions, using every tool at its disposal, including technology and foreign visitors (January 2022 Communications Editor's Letter). Considering how we have come to accept Chinese government atrocities in Tibet, you were at least honest enough to remind us of how Hong Kong is being subjugated, to the detriment of diversity. As reported in the New York Times (Jan. 30, 2022): "The Chinese Communist Party has found the best model for controlling people," "honed its powers to track and corral people, backed by upgraded technology …". "This amounts to a potent techno-authoritarian tool."

I would, however, contest the assertion "The situation is not one-sided" as contrasted with China's "… stark undermining of Hong Kong's basic law, and the violent suppression of its free press …" versus "the U.S. government has taken actions that discourage collaboration with Chinese institutions and researchers. Companies have been put on the entity list, blocking contact, and universities labeled as having military ties, limiting their students' study in the U.S." It is a matter of China's policy, practice, or advocacy of extending power and dominion versus the U.S. defending itself from attack and infiltration.

I do agree that it is a great shame that technological exchanges between countries, industries, universities, and researchers cannot be neutral and that we must guard against spying, theft, and attack. We went through this with Russia during the Cold War, but now the opposition is much more formidable.

Warren Scheinin, Redondo Beach, CA

Back to Top

Editor-in-Chief's response:

Thanks for your response. I raised the topic because thoughtful engagement is essential as too many in our community (the international ACM computing professional community) deny the existence of a growing problem. Or even if they recognize and acknowledge it, remain unwilling to engage to manage it. Only with the engagement of the bulk of our membership will ACM be able to address and surmount this rapidly growing challenge and forge a stable international collaborative community. We need the creativity and passion of the entire community, even those denying an issue now to help us find innovative solutions.

More directly, I believe you are mischaracterizing my comments. I called out actions from both the PRC and U.S. because indeed both governments are taking actions that erode our open, collaborative community.^a In no way did I suggest they are equal, nor did I characterize who might be viewed as an aggressor or defender. I leave it to others to engage in such discussions, as my focus is on how all of these actions erode trust, communication, and the ability to collaborate. This is to the detriment of computing science, education, and the computing profession.

Andrew A. Chien, Editor-in-Chief, Communications of the ACM, Chicago, IL

Back to Top

Systems and Methods to Detect Patent Vampires

In his November 2021 Kode Vicious column, "Patent Absurdity," George Neville-Neil advises software developers to avoid reading patents. He also recommends disallowing software patents, suggesting the objective of patents is to serve as weapons to achieve financial benefit and that they are contemptible and useless, proposing interacting with lawyers is like interacting with vampires, and not the friendly kind.

First, there is no such patent type as a "software patent"—this designation does not exist. There are inventions that can be implemented by using software; however, the inventions are processes and methods, instructions that can be implemented to create useful functionality. Second, many inventions give their readers the ability to quickly become exposed to novel concepts that are often easy to understand; for example, a new authentication method that is fundamentally different than all existing techniques, such as standard and biometrics-based passwords. Other inventions quickly expose the reader to novel computational methods; for example, a new feature selection method that competes with widely used techniques, all attempting to achieve better prediction and classification performance. Indeed, patent claims are often written in a legal language that is not easily understood. However, the content and figures are often readable even by nontechnical individuals. Such content is different from scientific manuscripts, which usually require significant expertise to understand in full.

Why does Neville-Neil express such negative opinions about software-related patents? This is likely because of a minority of people who abuse the patenting infrastructure. The U.S. Patent and Trademark Office as well as lawyers (the good kind) have taken a variety of measures to prevent issuing patents for obscure and/or abstract concepts. Furthermore, many of the people involved in patenting are legitimate lawyers and patent agents who truly wish to protect the novel, useful, and valuable inventions of innovators at the company they work for—and there is nothing wrong with that! An invention that describes a novel computational technique may be formed by a group of scientists and engineers working together for hundreds of hours. They and the company they work for deserve to financially benefit from their efforts.

What about this minority of people who abuse the system? Like how inventors form a novel system or method, legislative decision makers need to continue working toward creating more efficient methodologies to detect and then quarantine all patent vampires.

Uri Kartoun, Cambridge, MA

Back to Top

Author's response:

If only it was a minority that abused the system. The stream of cases passing through one, very famous, court in East Texas seems to show that the problem is large, and with another court in West Texas recently sending its own "welcome to patent trolls," perhaps to compete with East Texas, the problems will only grow; see https://bit.ly/3H0yig3

George Neville-Neil, Brooklyn, NY

Back to Top

Census Reconsiderations

Garfinkel, Abowd, and Martindale in "Understanding database reconstruction attacks on public data" (Communications, March 2019) highlight the dangers of database reconstruction attacks (DRAs). These U.S. Census Bureau methodologists illustrate DRAs using a small dataset of seven records, which they claim to be realistic because "the 2010 U.S. Census contained 1,539,183 census blocks in the 50 states and the District of Columbia with between one and seven residents." They protect the statistical output released for this dataset using cell suppression, more precisely "the rule of three," according to which cells sourced from fewer than three individuals are suppressed. The protected output is shown in Table 1 of their article, where (D) stands for a suppressed cell.

They show the original dataset can be uniquely reconstructed from their Table 1 by using a sophisticated SAT solver—in fact, reconstruction is feasible with much less apparatus: differencing entries in Table 1 and using simple logic and arithmetic reasoning success (see Muralidhar and Domingo-Ferrer, "Database reconstruction is very difficult in practice." UN-ECE/Eurostat Work Session on Stat. Data Confidentiality 2021.) Based on their analysis, they conclude protection of databases using suppression is inadequate. It turns out the reconstruction of Garfinkel et al. is based entirely on the incorrect application of (primary and complementary) suppression, so this conclusion is completely unwarranted:

• Under the rule of three, in addition to suppressing cells sourced from fewer than three individuals (primary suppressions), cells whose value would allow deriving the primary suppressions via subtraction must also be suppressed (complementary suppressions). See U.S. Census Bureau documents "American Community Survey: Data Suppression" and "Disclosure Avoidance and the Census." This requires cells in line 4A of Table 1 to be suppressed. Failing to suppress 4A allows differencing 2C and 4A to obtain 4B (the values for the only Black American male individual in the dataset).
• Releasing the median Age discloses the age of a single individual (when the group size is odd) or the average age of two individuals (when the group size is even). Hence, medians cannot be released under the rule of three.
• In contrast, there is no need to suppress the count and the mean in line 3A, since they affect three individuals (Single Adults) and do not need to be eliminated as complementary suppressions.

Without 4A, with 3A and without medians, there are thousands of datasets compatible with the remaining counts and means. No unique gender assignment to the seven individuals is possible. Also, uniquely reconstructing Age values without the medians is infeasible, as there are many Age assignments that fit the extant means and counts.

Hence, correct application of cell suppression can perfectly protect the example dataset. What the authors have proven, albeit unintentionally, is that simple disclosure prevention techniques, properly applied, are effective in preventing reconstruction of even very small (hypothetical) datasets created by senior U.S. Census Bureau methodologists to show the very opposite. Database reconstruction may not be so easy after all and this must be taken into account in the ongoing process of protecting the 2020 Census outputs.

Krishnamurty Muralidhar, Norman, OK, and Josep Domingo-Ferrer, Catalonia

Back to Top

Authors' response:

We thank Muralidhar and Domingo-Ferrer for their contribution to the example, and we note that the Census Bureau chose to use differential privacy rather than the alternative, massive cell suppression for the 2020 Census.

Simson Garfinkel, Arlington, VA, John M. Abowd, Washington, D.C., and Christian Martindale, Durham, NC