HNHacker News
TopNewBestAskShowJobs

djao

145 karma · joined September 5, 2012

submissionscomments
djao··on I drove hundreds of miles hands-free in GM, Ford, Tesla cars here’s how it went
1. No current system is truly hands-free. The GM and Ford systems technically allow you to take your hands off the wheel, but this compromises safety. Tesla could do the exact same thing (they have an in-cabin camera which can tell if your eyes are paying attention to the road, same as GM and Ford). But they don't, because it's unsafe. This isn't because the Tesla system is less capable or less safe than GM/Ford. (It's much more capable, and much safer.) It's because Tesla's safety standards are much higher. The article doesn't mention this.

2. The GM and Ford systems only operate on pre-mapped highways, and were (obviously) only tested on pre-mapped highways. The Tesla system was given only a "limited test" on the highway. Essentially, the article compares GM/Ford's highway driving to Tesla's city driving, while barely mentioning that GM/Ford can't drive in the city at all (whereas of course Tesla can easily do highway driving). This is a ridiculously biased comparison, crossing the line into fraud. An honest journalist could easily have compared highway driving to highway driving. The obvious conclusion is that this journalist is dishonest.

3. The article contains glaring factual inaccuracies. Among the worst is the claim that "the only indication that Tesla's ADAS is engaged is a tiny steering wheel icon - smaller than a dime". This is flat-out false. There is a prominent and reliable audio cue that tells you whenever the system is engaged or disengaged. You can look at the color of the trend line in the car visualization display. The gear display changes color, as does the speed limit display. If nothing else, you can torque the wheel (you are keeping your hands on the wheel for safety, right? Right????) and feel the level of resistance.

djao··on The post-quantum future: challenges and opportunities
> Most PAKE systems nowadays use Diffie-Hellman assumptions, which can be broken by a quantum computer. There are some ideas on how to transition their underlying algorithms to the post-quantum world; but these seem to be so inefficient as to render their deployment infeasible.

Isogeny based PAKEs exist! (e.g. https://eprint.iacr.org/2018/886) Their weakness is not infeasible performance, but rather the lack of a totally satisfactory security proof based on previous assumptions.

djao··on Freenode IRC resign en masse after takeover by Korea’s crown prince
This is absurd. The timeline is irrelevant. Andrew is committing gross abuse of power. Regardless of why or how such abuse is "justified", the abuse in and of itself is reason to leave the network.
djao··on Freenode ops take control of 700 channels
If WhatsApp users want to switch en masse to a new platform, they have to manually reconstruct their chat groups and their contact lists, install an entire new app, learn how to use the new app, work around any features that WhatsApp had which are now missing on their new platform, and probably a bunch of other stuff I'm forgetting.

If IRC users want to switch to a new server, just use the /server command. Done. Maybe you have to re-register your nick or channel. But the client is the same, the interface is the same, everything stays the same.

djao··on Don't Become a Scientist (1999)
In 1980, about 2% of R01 recipients were under age 30. So yeah, it's early, but not totally impossible.

In 2010, about 0.1% of R01 recipients were under age 30. That's bordering on impossible.

djao··on Salted Password Hashing – Doing it Right
The part about having to use HMAC for keyed hashing is (hopefully) going to be obsolete soon. SHA-3 is explicitly designed to be safe under the H(key+message) usage.
djao··on Show HN: Check if your browser is vulnerable to the Apple SSL bug
That makes it slightly harder. Still, Apple will sign anything that is successfully submitted to the App Store for approval, right? So you just need to slip a single trojan past their approval process. Normally, the transport layer provides an important second level of defense: a victim would have to consciously choose to install YOUR program in order to get hacked. That doesn't work anymore when the security of the transport layer is compromised. Now anything that you install from the App Store could be surreptitiously compromised. Better hope that Apple's signature revocation infrastructure is sound.
djao··on Show HN: Check if your browser is vulnerable to the Apple SSL bug
Doesn't matter, the system updater can blow away Chrome or anything else.
djao··on Show HN: Check if your browser is vulnerable to the Apple SSL bug
Anyone who thinks Chrome and Firefox are safe from this bug doesn't understand the issue. SecureTransport is used for updating software. So an attacker could trick you into installing a malicious update to Chrome, FireFox, or for that matter anything on your system. They could even slip in malware under the guise of a patch that purportedly fixes this bug. Using alternative browsers does NOT completely protect you.
djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
Generally speaking, attacks which rely on more structure are hard to adapt to situations where there is less structure. Conversely, attacks which require no structure are easy to generalize to a wide variety of situations.

Something like Pollard rho requires no structure. It works in any group. The number field sieve is an intermediate case. It requires a norm (basically you need some elements to be bigger than others). For arbitrary finite fields, the NFS still represents the fastest known attack. The latest attack is a version of the function field sieve. The FFS uses, in an essential way, the algebraic structure of polynomials over a field. Coppersmith pioneered the use of the FFS in the 80s, when he showed that small characteristic fields (which consist primarily of polynomials) admit faster discrete logarithm solutions. The latest work goes far beyond Coppersmith's work, but tends towards increasing specialization, not generalization, since it adds the new requirement that the field contain an intermediate subfield.

Given that the direction of specialization appears, if anything, to be going in the other direction, there would have to be a substantial further breakthrough in order to apply these results more generally. My own opinion and intuition is that these results are not likely to apply to prime field discrete logs, never mind RSA and/or composite moduli.

djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
The thing is, some people think we're already on a 10-year clock. Quantum computers represent a very real threat, and there are no theoretical obstacles to their implementation. So yes, we should be migrating to something else. These latest breakthroughs add more impetus, but I don't think they're going to get there before quantum computers (just my opinion).
djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
My experience as someone who is active in research in this exact field (mathematical cryptanalysis) is that generic techniques for discrete logarithms, such as Pollard rho or the number field sieve, do transfer well to integer factorization, but more specific techniques such as Joux's latest function field sieve tend to be harder to transfer from one setting to the other.
djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
Attaching specific numbers to an expression like L(1/4) is meaningless, because L(1/4) is an asymptotic expression. Asymptotic expressions contain implied constants, whose values are not explicit. There's also no particular reason to expect the L(1/4) attack to apply to RSA. For one, I explained in another reply that we currently do not know any direct relationship between RSA and discrete log. In addition, Joux's latest result is not even L(1/4), it's L(epsilon) for any fixed epsilon > 0 (and in this context, the implied constant cannot be ignored, otherwise you get nonsense values).

Google does use ECC, on platforms that support it. See for example http://googleonlinesecurity.blogspot.com/2011/11/protecting-... which has a screenshot.

djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
There is a difference between "Here's an example of a cryptosystem that got attacked" and "Here's an example of an attack against RSA."
djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
You are perhaps referring to Boneh's comments at the RSA conference last February. Boneh cited the new attacks in the context of arguing that we should diversify away from RSA and DH. The implication, to me, was one of general caution and urgency. I did not get any sense of RSA itself being specifically threatened.
djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
Well, now you're changing the goalposts. You originally said 64kbit, not 16kbit. Key generation is roughly O(n^4) in the bit size. If you quadruple the key length, the time goes up by a factor of 4^4, or 256. Key generation time is not irrelevant -- waiting ten hours before your SSH server can come online is no one's idea of a good user experience.

At the 64kbit-level (again, BIG BIG difference from 16kbit), even the cost of encryption/decryption alone is easily going to cripple a high-traffic server.

djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
The only relation between the two problems is that they involve the same type of equation. The underlying number systems are totally different. There is no reason to expect the two problems to be of comparable difficulty.

As an illustrative example, the equation for the RSA problem itself can be defined over prime fields, and it is trivially easy to solve in that setting. The same problem over a composite modulus is believed to be hard, and forms the basis for the RSA cryptosystem. Just because an equation can be solved over prime fields does not mean, at all, that it can be solved over composite moduli.

djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
You're assuming unrestricted discrete logarithms over a composite modulus. This is a much stronger assumption than discrete logarithms over prime moduli. Even if someone found a fast way to break discrete logarithms modulo primes, your assumption would still not automatically hold.

Since cryptography uses discrete logarithms over prime moduli, the discrete logarithm problem that cryptographers are working on is not the same problem as the problem that they would need to solve in order to break RSA.

djao··on Math Advances Suggest RSA Encryption Could Fall Within 5 Years
RSA with 65536-bit keys is quite secure, but it's hugely impractical. Try it in OpenSSL. Key generation alone will take you several hours on a fast, current-generation machine.
djao··on Shin Mochizuki has released his long-rumored proof of the ABC conjecture
Also among the blog commenters is NDE (Noam Elkies), who was my Ph.D advisor.
← PreviousPage 4 of 4