Orr Dunkelman

CS, Cryptography, Cryptanalysis, Privacy, Biometrics, Computer Security. (personal account at @mkilmo). He/Him.

Also @[email protected]

[Orr Dunkelman @CryptoOrrDun]

I love the smell of cryptanalysis in the afternoon!

[Orr Dunkelman @CryptoOrrDun]

First Ph.D. student of mine to land an academic position.

May many more join.

Good luck Shibam!

[Shibam @Vola_Shibam]

I am delighted to share that I have joined the Department of Computer Science and Engineering at IIT Hyderabad as an Assistant Professor. Thanks!

[Orr Dunkelman @CryptoOrrDun]

@Khovr to be used, without me getting the credit I deserve for it.

And I have many more criticism about LLMs.

So I don't mind the luddist title.

[Orr Dunkelman @CryptoOrrDun]

I read eprint 2026/1035, so you wouldn't have too. Independent of my thoughts about standardizing Rijndael-256, I believe that this paper should be rejected from any venue for being AI slop (and a bad one at that).

1. It is a badly written SoK. Many recent results are missing >>

[Orr Dunkelman @CryptoOrrDun]

@Khovr This is a more delicate discussion than twitter allows. But when an LLM does a "research" can you attribute the previous research it is based on? LLMs do not know (and it seems unlikely they can really explain) what they are basing their facts upon.

So, my paper is going >>>

[Orr Dunkelman @CryptoOrrDun]

@Khovr I disagree. But I don't mind being called a luddist once every while.

[Orr Dunkelman @CryptoOrrDun]

Given that everybody are now vibe coding, I believe that attacks a-la the ones reported in Ken Thompson's seminal "Reflections on Trusting Trust" will become easier to execute and harder to detect.

[Orr Dunkelman @CryptoOrrDun]

@Khovr AI should be allowed for "style" touches. Fixing your English because it is not your native language. Making paragraphs clearer.

They should never ever be allowed in papers (even SoK) for anything above that.

[Orr Dunkelman @CryptoOrrDun]

@pratiks_crypto The problem:
They wasted 10 minutes creating a paper that it would take me two full days to dismantle. While these two days are community work and service for the better good, I also have to do other things to write something which is sufficiently accessible and clear.

[Orr Dunkelman @CryptoOrrDun]

@MMahzoun Remind me - is there an intersection between the authors' list?

[Orr Dunkelman @CryptoOrrDun]

Now, if you want to argue for Rijndael-256 being standardized - be my guest (I have my views, and I've expressed them in some forums).

But please be accurate in your "security" analysis, and don't AI slop arguments that most people can't tell if they are true or not.

[Orr Dunkelman @CryptoOrrDun]

And don't get me started on Table II which is a weird combination of semi-facts and sort-of-(either misunderstandings or hallucinations).
>>>

[Orr Dunkelman @CryptoOrrDun]

@veorq I see that the cats jumped out of the box... eh... book.

[Orr Dunkelman @CryptoOrrDun]

@veorq Oh yes!

[Orr Dunkelman @CryptoOrrDun]

Tokens for AI is what happens when people who grow up on microtransactions and pay for win build payment systems and options.