Post AIyxKDkLeDB8QFvSTY by udunadan@ioc.exchange
 (DIR) More posts by udunadan@ioc.exchange
 (DIR) Post #AIyt45Gh0D7lg46Xh2 by udunadan@ioc.exchange
       2022-04-30T16:57:46Z
       
       0 likes, 0 repeats
       
       The amount of papers I've been critical of at first and forgot to check on them later due to delays in source code publishing to verify the claims made is staggering.
       
 (DIR) Post #AIyt45j3IloF61NBtg by seb@ioc.exchange
       2022-04-30T17:00:09Z
       
       0 likes, 0 repeats
       
       @udunadan What area of study/research?
       
 (DIR) Post #AIyuyHOXCfZBSS8HIW by udunadan@ioc.exchange
       2022-04-30T17:21:35Z
       
       0 likes, 0 repeats
       
       @seb cs.CR (cybersecurity)
       
 (DIR) Post #AIyx37ACryR35Crcye by udunadan@ioc.exchange
       2022-04-30T17:29:01Z
       
       0 likes, 0 repeats
       
       @seb but actually, a lot of intersections were present, ML most of all: it's always interesting to evaluate the claims of its applications to security problems and often the source/data materials are to be disclosed later and sometimes they do not publish stuff at all.
       
 (DIR) Post #AIyx37nuUMC74LRKrY by seb@ioc.exchange
       2022-04-30T17:44:50Z
       
       0 likes, 0 repeats
       
       @udunadan I have been a little cautious about the ML approaches lately - in blue team / defense they don’t always work as nicely as people think.While cybersecurity is a numbers game to a certain degree, it also pivots at a fast pace. I’m unsure whether it is possible to adjust ML models at the same rate.
       
 (DIR) Post #AIyxKDkLeDB8QFvSTY by udunadan@ioc.exchange
       2022-04-30T17:47:58Z
       
       0 likes, 0 repeats
       
       @seb I think that the computational spaces where the output function is binary (simply speaking, we need to answer "yes" or "no", like with malicious behavioral analysis) ML fits nicely and even if it works bad now, it will become better in future. But attempts to plug it into every possible usecase are not very promising.Still, we shouldn't pre-assume stuff and we better evaluate it scientifically. For that, we need research transparency & standards to be held.
       
 (DIR) Post #AIyxtmHtDVaRgWZMkS by seb@ioc.exchange
       2022-04-30T17:54:21Z
       
       0 likes, 0 repeats
       
       @udunadan Agreed. Time will tell.