Posted by: jonkatz | August 12, 2009

Asiacrypt 2009

Notifications for Asiacrypt 2009 went out this week. (The list of accepted papers is not public yet; I will post a link once it is.)

Feel free to gripe about any rejected papers here. =)

Apparently, competition was really tough this year. I heard from more than one person that good papers (in one case, a paper I thought was a definite accept) were rejected. (I had a paper with three very positive reviews and one neutral review rejected.) In an email, the program chair said that there were roughly 300 submissions, and so many good ones that the committee decided to accept more papers than usual (42 as compared to 33 the previous year), even at the expense of giving up a slot for an invited talk. Even with this, the acceptance rate was a paltry 14%.

I’m having a hard time figuring out why there were so many (good) papers submitted. Crypto this year had roughly 225 submissions, and if I recall correctly Eurocrypt had around the same or slightly lower. Given my earlier post arguing that the crypto community publishes too many papers, and that we have too many conferences (both of which I still believe to be true), where did all these (good) papers come from?

It’s fair to say that, although Asiacrypt is one of the “flagship” IACR conferences, it has a reputation as being a tier below Crypto/Eurocrypt. Over the past few years, however, I think Asiacrypt has become much stronger while Eurocrypt, in particular, has become a bit weaker. (Is the high submission count to Asaicrypt a reflection of this?) This is just my own opinion; I’d be interested to hear what others think.

The problem for me with Asiacrypt is that it’s too difficult/expensive for me to travel to it. I have never previously attended Asiacrypt, though I am hoping to this year.

Addendum (Aug. 15): Looking at the stats here shows that the submission numbers were literally off the charts: 300 submissions is close to the all time high (for Asiacrypt) of 314 submissions, and in no other years did Asiacrypt have more than 240 submissions. The all-time highs for Crypto and Eurocrypt are 220 and 206 submissions, respectively.

One more observation: Eurocrypt submissions this year were historically low, and I think this was in part due to the very early submission deadline for Eurocrypt this year. So maybe some papers that (in other years) would have gone to Eurocrypt were sent to Asiacrypt this year instead.

Advertisements

Responses

  1. It seems like every year one can find many accepts and rejects from Asiacrypt that are obviously very wrong when the papers are juxtaposed. How does this keep happening?

  2. 1) I was impressed by quite a few of the papers at Eurocrypt this year. They made my trip there worthwhile, even though I stayed only for two days (I had to return to teach a class…). So kudos to the committee (I’m looking at you, Jon) for overcoming what were apparently difficult working conditions to come up with a decent program.

    2) One obvious question: is the trend of increased quality at Asiacrypt (if it isn’t just a fluke) due to the fact that more better research is now going on in Pacific Rim countries, notably in China? Are there statistics on where the surge in submissions is coming from? The accepted papers list might give some indication, once it is published.

    3) For any given conference, one can find pairs of accepted/rejected papers that got classified the “wrong” way (call them unstable matches… based on the beholder’s preferences). In fact, even committee members usually have their own pet pairs like this. I don’t know if Asiacrypt — or Eurocrypt –has an unusually high number of weird decisions. Nor can I think of any reasonable way of testing this hypothesis…

  3. It is actually an interesting question why Asiacrypt got so many good submissions, may be it’s finally moving up to the same level as Crypto and Eurocrypt in people’s conscious? Note that *formally* it is one of the three IACR flagship conferences.

    Also, I think this gives even more weight to the need of somehow “marry” different conferences. If there were really many “good” papers that were rejected, then if all those authors would continue to polish their papers for the next Eurocrypt, while all the Eves and Mallories have had a few extra months to polish and write up completely new papers… It’d be a tough call.

    (Disclaimer. I also got an (imho) excellent paper rejected.)

  4. It seems like every year one can find many accepts and rejects from Asiacrypt that are obviously very wrong when the papers are juxtaposed.

    This problem is not specific to Asiacrypt, of course…

    Is the trend of increased quality at Asiacrypt…due to the fact that more better research is now going on in Pacific Rim countries, notably in China?

    Interesting. How would one test this? Note that Asiacrypt always gets a large fraction of papers from Asia.

    Flipping things around, I wonder if more people from the US/Europe submitted to Asiacrypt this year than in years past. That could indicate an increase in the the community’s perception of Asiacrypt.

    If there were really many “good” papers that were rejected…

    I’m still amazed by the idea that there can be so many good papers. Are we measuring “good” in an absolute sense or a relative sense. I.e., are these objectively good papers for which there is a compelling need to have them published? Or has the bar just been lowered so that we define “good” by the weakest paper accepted to the last conference? (Helger: I’m not taking about your paper here, of course. =) )

  5. It would be interesting to know the number of (accepted) papers, when excluding those directly related to the SHA-3 competition.

  6. 1. I guess many people chose to submit to AC this year because it was in Tokyo. I’m curious how the quality is this year. I already already heard of a couple of excellent papers that got rejected with terrible reviews.

    Btw, the locationmay also be the reason why ICALP Track C died out since the 2007 location was not too exiting compared to Iceland (2008) or Rhodes (2009). PKC 2008 had a low number of submission since it was in Barcelona only a couple of month after EC was there. Knudsen has a webpage with the number of submissions/accepted papers for our top conferences:
    http://www2.mat.dtu.dk/people/Lars.R.Knudsen/accrates.html

    2. @Anonymous: why do you think is the number of paper related to SHA-3 is relevant?

    3. I think in general we have to think about ways to deal with this high (>200) number of submissions. Otherwise the results will essentially become random.
    The chairs of ACM Sigcomm 2009 had some interesting ideas on which they report here
    http://www.pittsburgh.intel-research.net/~kpapagia/sigcomm09/

  7. I see at least an extra reason justifying the high number of submissions: At least from Europe (and the US West Coast) reaching Tokyo (the location of Asiacrypt 2009) is much easier (and cheaper) than it is the case for other locations were Asiacrypt was previously held (not surprisingly, something similar happened with the one AC taking place in Shangai).
    Last year’s AC in Melbourne had quite lower submission numbers, and, as far as I remember, some authors did not make it to give a talk there.

    Add, on top of this, that hash function crytpanalysis is booming (for obvious reasons).

  8. Jon, I agree, it’s outrageous they rejected 1 of your papers and accepted 4 :).
    Congratulations!
    Speaking of which, why did you send so many nice papers yourself, if you think AC is second tier?
    Tokyo? Maybe you can answer your own question…
    Yevgeniy

  9. For me personally (and I guess for many in Europe) it takes considerably less time (and money) to fly to Tokyo than to Santa Barbara. More precisely, if’d pay for a round trip right now, I could go to Tokyo in December for 640 euros, and to Santa Barbara for 1130 euros. I can also get to Tokyo in 12 hours, while getting to SB takes considerably more time (but it just might be my bad luck).

    Plus, don’t forget, there are many cryptographers *actually* living in Asia, especially in Japan, and Australia. Top people are still based in Europe/Israel and NA, but Asia is slowly catching up.

    I would also guess the SHA3 competition played some rôle.

  10. Aren’t the following obvious reasons good enough?

    1. The good papers that were borderline cases in Eurocrypt/CRYPTO and were rejected got submitted to Asiacrypt. (The number of rejected papers in EC+CRYPTO ~ 115+175).

    2. The good papers that were deemed by the authors not to be competitive enough (but good quality nevertheless) to make it to EC/CRYPTO got submitted (only) to Asiacrypt.

    3. There number of cryptographers around the world (esp. in Asia – which makes traveling to Asiacrypt very feasible) is increasing every year.

    4. Work during the winter (the no-vacation period) got submitted to Asiacrypt because of the May deadline. =)

    Note that since (1) and (2) above are one-way (at least as long as AC is classified as a second-tier conference), the same phenomenon would repeat itself.

  11. Ah I see… so the question is why so many submissions to AC *this* year? I just looked up http://www2.mat.dtu.dk/people/Lars.R.Knudsen/accrates.html
    for submission stats for previous years…

  12. Speaking of which, why did you send so many nice papers yourself, if you think AC is second tier?

    There were several factors:

    (1) The location definitely helped. While it will be difficult for me to get to Tokyo, it is more likely I can/will go there than some other places Asiacrypt has been held in the past.

    (2) There was also an issue of timeliness. For papers that were ready months ago it didn’t feel quite right to wait for the TCC/Eurocrypt deadlines. (On the other hand if it turns out that I can’t go to Asiacrypt, this will end up being a bad decision…)

    (3) I was on the committees for both Eurocrypt and Crypto, and this created somewhat of a backlog.

  13. I wonder how the situation in crypto compares to that in algorithms in terms of the paper acceptance rate. Here is a question: There is the SODA conference which, as of last year, had about 450 submissions and 135 acceptances. So they have about a 30 percent acceptance rate. Is this different from having three conferences, (crypto, EC, and AC) each with a 30% acceptance rate, each accepting 30 papers? I think so, since in the crypto scenario papers can be submitted more than once leading in an overall acceptance rate of more than 30%.

    So in answer to the question of why so many conferences? one reason is that many conferences can have the effect of making it look like a field is really competitive with low acceptance rates, but in fact be not that competitive with higher overall acceptance rates.

  14. asterix makes an interesting point. There is some truth to it, but let’s not forget STOC/FOCS. Each accepts 75-90 papers (say, 80 average), among which usually 5 are crypto-related, and could appear at CRYPTO/Eurocrypt/TCC. Out of remaining 75, perhaps 60+ could easily appear at SODA (definitely 50+ to be cautious). Given the acceptance rate of STOC/FOCS is above 30%, this means overall acceptance rate of alg. papers is 30%, with 3 conferences, just like CRYPTO. Of course, this is not completely fair as well, but my point is that things are trickier, although there is SOME truth to your statement.

    Jon, don’t serve on these many committees :)!
    Yevgeniy

    PS. come to Tokyo, I’ll be there, even without any papers :).

  15. Tokyo is fun. I hope Yvo will shows us around there.

  16. Accepted papers are online:
    http://asiacrypt2009.cipher.risk.tsukuba.ac.jp/program/index.html


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: