Last year a few good friends and people from our discord server started up a ‘Paper Group’ where each week a new research paper is selected for us to read and discuss. We all found it to be an enjoyable experience & are continuing it indefinitely. This page will covers the papers selected for 2022, some of our comments, and at the end of the year we’ll include highlights such as best and worst papers of the year, etc.
- Is Justified True Belief Knowledge? – https://fitelson.org/proseminar/gettier.pdf
- This brief paper discusses what it takes to claim one has knowledge of something. It provides a strong case that ‘a thing being true’, ‘belief that the thing is true’, and ‘justification for that belief’ are not sufficient in and of themselves to constitute knowledge.
- The discussion was really enjoyable and we talked about how either expanding the meaning of ‘justified’, adding additional criteria, or replacing justification might lead to a workable definition of knowledge.
- There was also good discussion about why or if the topic – ‘what is knowledge’ – and why it may be worth discussing. We touched on how the value of a standard criteria to evaluate if someone knows something is useful socially and legally.
- I highly recommend thinking about this paper a little bit, it’s also a great example of how a short two page paper can challenge beliefs and lead to decades of discourse searching for a better criteria.
- Obfuscation Revealed: Leveraging Electromagnetic Signals for Obfuscated Malware Classification – https://dl.acm.org/doi/abs/10.1145/3485832.3485894
- This paper presents a ‘new’ approach for detecting malware and obfuscated malware by using EM signal monitoring to fingerprint behavior types.
- It starts by tearing down other approaches as insufficient which seems largely valid as detection is still poorly done across the industry. It then focuses on ARM IoT devices as their target detection goal; primarily because security is less emphasized in design/tooling of these products & there are so many so it poses as broad attack vector.
- It seems to be very impressive at detection but very curious how the EM monitoring of the CPU was done (e.g. specific parts of the CPU?) and how advanced the detection tools need to be.
- Some concern that the detections were not ‘real-world’ enough and simply looked at individual program executions and compared complex things (malware) to simple binaries (ls) and the detections were perhaps oversimplified? That being said 99.82% on the type of malware is a reallly good number; very curious to see follow-up work on this paper.