How Hard Is It to Detect Surveillance? A Formal Study of Panopticons and Their Detectability Problem

Round 1

Reviewer 1 Report

Liagkou et al. present a work where they try to combine massive surveillance methods with formal techniques from theoretical computer science and try to prove that even the detection of those privacy-violating methods is impossible (in terms of decidability results from a Turing machine operation).

The main idea behind the paper seems novel at first and I particularly like the theoretical insights it may offer as a broad topic (but not the results therein). But I have some scepticism over the merit of the paper and it message.

 

The introduction needs revisions. The biggest part of it looks like a pop science article rather than a formal science paper. And this is evident when one looks at the second part of it, where the paper's contribution is summed up. Terms and concepts from computation theory appear out of the blue and that's where for the first time we realise the paper's most important weakness: the correlation between the theoretical Panopticon and formal methods. And this is significant because it's actually the research question the authors pose and try to ask here. In my opinion, that exact weakness is crucial and sabotages the paper's soundness.

After the standard definition of Turing machines, some bizarre ones follow, like the so-called behavioural Panopticons. No matter how much I tried, I failed to understand what they correspond to. The whole idea seems like transforming a Panopticon-like model to an input for some Turing-like machines, and subsequently trying to prove some decidability results upon them. But, in reality, I can't see why Panopticon are unique in that respect. 

 

One minor observation I made is the overuse of commas throughout the text. This of course by no means distorts the paper's scientific merit, but still hugely affects the paper's readability at some points and confuses the reader (e.g., when splitting linked parts of speech).

 

Overall, in my opinion, this work is weakly founded with a lot of research abstractions here and there that blur the paper's narrative. Therefore, I cannot unfortunately recommend its publication.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript reports: The manuscript reports: How hard is it to detect surveillance? A formal study of Panopticons and their detectability problem. The manuscript it's interesting and I am expecting high interest of the readership. The article could be considered for publication after stronger revision along the suggestions. The abstract is infinite and qualitative, the authors should put the most relevant results. Several sentences used in the abstract would help improve the introduction. The introduction of relevant background and research progress was not comprehensive enough. Please carefully review the references and how they were used in the work. The introduction of the manuscript contains few references and without a logical order. It is appreciable that the authors have analyzed How hard is it to detect surveillance? A formal study of Panopticons and their detectability problem, but there is no in-depth scientific discussion. A careful review of the entire manuscript must be performed. Strong scientific discussions must be made throughout the entire manuscript to prevent the work from being overly qualitative. Compare your results with other works reported in the literature, show why your work is necessary and highlight contributions to scientific development. References must be updated, most are old, which influences the current work

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

 

In this paper, the authors present an study about he detection of “panopticons” with a notable formal approach. The paper is well written, and the structure is correct.

When I was revising the paper I was surprised because most of the references are more than 20 years old. Searching for more modern references I came across the paper entitled “On the Undecidability of the Panopticon Detection Problem” authored by the same authors (https://link.springer.com/chapter/10.1007/978-3-031-07689-3_6 and https://ui.adsabs.harvard.edu/abs/2021arXiv210705220L/abstract ). The present paper is a word-by-word copy of the archived technical report but for section 7 and two paragraphs of the conclusions section. Thus, my biggest concern, is about the actual contributions of the present work.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

First of all, I wish to thank the reviewers for taking into account my comments and providing a revised version along with their responses. It is clear that the manuscript has improved and the message is now clearer. I have to say that I'm still sceptical over the level of contribution, but I do not want my personal opinions to become an obstacle to the publication path of this research, especially after the first revision round. The paper is now more solid from a scientific perspective and I think readers within the cryptography and (mostly) the privacy community might find it useful. Below I have some minor comments and a short list of typos.

 

A quick note for future works: transforming the so-called Panopticon's behaviour as an input to a state machine seems to share a lot with some variants of Markov decision process, namely those with hidden states. Like, for example, the passive actions "observed" from the operation of a Panopticon could correspond to hidden states in POMDPs?

Minor comment:
When defining the different Turing machines, try to follow the tuple sequence when expressing the meaning of each element. That is, if the tuple is (A,B,C,D), then it would be better to explain A first, then B and so on. This will help a lot to easily distinguish between the different TM variants defined throughout the manuscript.

 

Typos:

line 60: Turing Machine based -> Turing-Machine-based

line 77: too -> to

line 85-86: can, not only cannot

Reviewer 2 Report

There is clear evidence of manuscript improvement. I consider that this version can be accepted for publication

Reviewer 3 Report

The authors have explained that this is an extended version of a short paper presented at the CSCML 2022 conference. This is something that we reviewers do not know… If this extended version meets the requirements of the journal is something that we cannot fully check because the proceedings are not publicly available. Nevertheless, as I stated before, this paper is a word-by-word copy of the document in arXiv but for sections 7 and the conclusions (partially). If this is customary or not is something I cannot tell, but I can confirm that this may create copyright conflicts. If the editor agrees with the situation, then I find the paper good enough to be published.