Breakfast Bytes Blogs

The biggest security conference in the world is RSAC, held in San Francisco. It fills all three buildings of the Moscone Center. Some of the monitors around display some interesting statistics, such as the over 400 exhibitors shipped over 225 tons to the show. There were over 26,000 attendees and over 600 speakers in over 350 sessions.

If RSAC is the highlight of the cryptographic year, the highlight of the conference (in my opinion) is The Cryptographers' Panel. This year the panel was:

• Adi Shamir, the Borman Processor of Computer Science at the Weizmann Institute in Israel. He is also the S in RSA, the encryption algorithm on which internet security depends.
• Whitfield Diffie, whose title is Cryptographer and Security Expert, Cryptomathic. He is also one of the two authors of the original Public Key Cryptography, also known as Diffie-Helmann.
• Dawn Song, Professor of EECS, Director of Center on Responsible Decentralized Intelligence at UC Berkeley, here in California.
• Moni Naor, Judith Kleepman Professional Chair Weizman Institute of Science in Israel.
• As usual, the panel was moderated by Zulfikar Ramzan, the Chief Scientist at Aura and CEO of Aura Labs (and check out those RSA shoes)

The Mathematics Prize

Just before the Cryptographers' Panel was a presentation of the Award for Excellence in the Field of Mathematics, which is co-sponsored by The International Association in Cryptographic Research. Normally, Ronald Rivest presents this award, but this year he could not make it, so Adi Shamir presented it. The honorees were:

• Cynthia Dwork, who is the Gordon McKay Professor of Computer Science at Harvard University.
• Moni Naor (see above since he was on the panel)

Cynthia was unable to attend and shake Avi's hand and receive her award. You might have noticed that both Avi Shamir and Moni Naor are at the Weizmann Institute. In fact, their offices are next to each other, meaning, as Avi pointed out, they both had to fly halfway around the world to stand on the stage and shake hands instead of just going ten yards down the corridor.

The Panel

To keep things clear, I'll indicate the questions, which means they were actually asked by Zulfikar. Also, anything in [brackets] is my commentary and not something said by the panelists.

Q: NFT, where are we with them? [NFT are Non-Fungible Tokens] Moni, you invented the idea of proof of work protocols that have been the foundation for many of these technologies.

Moni: The most interesting issue is where do you store what you are buying. You can't store it on the chain, it is too big.

Adi: The numbers are down by 90% in just a few years. There was the recent case where an NFT was purchased for $29M, attempted to sell it for $45M, but the highest bid was $45,000.

Q: The word "crypto" has been co-opted. Maybe with the dwindling process, we can recapture the term for cryptography. Let's switch to fungible tokens, such as cryptocurrencies. What do you see as the challenges in this space?

Dawn: The big challenge is scalability. How can we develop scalable infrastructure with decentralized tasks? People have heard about Ethereum, so this is just one datapoint. Uka Labs sold NFTs worth $300M, but $180M was burned just to support the transactions. We need new techniques to build scalable infrastructure. In the crypto world, losses are real. Just this year, financial losses caused by attacks in the crypto world is over $1B. The London Bridge attack alone caused over $600M loss. We need to enable provable guarantees and privacy. People think Bitcoin is private but actually only pseudonymous.

Q: What about mobile? Passwords? Biometrics?

Whitfield: I am the world’s last fan of passwords. Security officers like other things such as biometrics since it gives them more ways to control, which is what they want. But passwords are something you know. So you have a sort of 5th amendment right not to tell anybody. I lack confidence in biometrics since there might be something suddenly discovered that has a huge failure. At an airport, looking at their eyes, there is not so much. But at home maybe can have an artificial eye. This is password's last stand.

Q: “If something possibly can go wrong, it will.” How do you implement cryptography and guarantee it is correct?

Adi: The number of zero-day attacks [a "zero-day" is a vulnerability that was completely unknown until it was suddenly used]. One area that is most affected is mobile security. There were a couple of companies, one in Italy, that are now on the blacklist of the US government and since then have had no sales. But there is no vacuum. Other companies will come to the fore.

Q: If someone can build a quantum computer at scale, two of you on this stage have a dog in that race [Avi Shamir is one of the co-creators of the RSA algorithm, and Whitfield Diffie was one of the co-creators of public key cryptography]. How do you see quantum computing evolving?

Adi: I have been following this very closely. The state of art is 50-70 qubits with no error corrections. That can be used for publicity stunts (like Google, which showed it couldn’t be done with a regular computer). At end of 2021 IBM showed a device with 127 qubits. Since then half a year has passed and nothing more has been published or announced. It is not clear if it is gearing up for a big announcement, or perhaps failing. Another big development at Microsoft uses topological qubits, and for about 10 years were unable to build even one such qubit, but a few months ago there was a breakthrough, but they still have not got their first real qubit.

Q: So good for RSA and Diffie-Hellman for now. Where are we in post-quantum, algorithms?

Adi: NIST [The US National Institute of Standards and Technology] started in 2017 to accept algorithms for post-quantum. There were 69 proposals, three rounds of analysis. Now in the third round it is down to three public key encryptions and three public key signatures. A month ago, an Israeli military scientist showed 3 of them have security issues that drop below NIST’s standards. Kyber, Saber, and Delithium. Security dropped by amounts up to 2 to the 14th, even though this paper was published by the Israel military and hard to publish since Arxiv won’t publish a paper without authors' names [anonymous].

Q: Let's complete the buzzword bingo with machine learning.

[Dawn Song's work on adversarial neural networks has appeared previously on Breakfast Bytes, see my post RSA Wrapup: Song, Darling, Thrun]

Dawn: First privacy-preserving machine learning, second adversarial machine learning. I am very optimistic about privacy-preserving ML, where we have made many advances such as fully holomorphic encryption [see my post Fully Homomorphic Encryption], and some of these have been integrated into things like TensorFlow. There is also hardware acceleration. Advances in hardware are required to enable privacy-preserving computation. I have a prediction: in 10 years secure computing—privacy-preserving computing—will be commonplace. Secure enclaves will become prevalent. I'm excited and optimistic that we will be in a different place, which will be great for the whole world. But on adversarial machine learning I am quite pessimistic. In last 4 years, there are thousands of papers published in the domain, but we have not actually made much progress, especially compared to privacy-preserving machine learning. In 20 years, I predict we will still be battling this challenge.

Adi: It is clear to all the practitioners in this field that neural networks are very fragile. For example, take a picture of a cat and change a few pixels, and now it is guacamole. People have been looking into how to stop these tiny changes from breaking everything, but it seems to be ingrained in the model of deep neural networks. This is very different from how the human brain works. A tiny kid will distinguish a cat from guacamole and will not be fooled by a few pixels. In the last year, people have published many papers showing that the three best-known facial recognition programs are fooled by “master faces.” 10 normal-looking faces will be wrongly recognized as being the same as 40% o the population on earth. If you take a person and one of those 10 master faces, they will be seen as the same.

Q: GDPR, where is that heading? [See my posts GDPR Starts Today and "GDPR Is an Enormous Regulatory Own Goal"]

Dawn: GDPR is the first step and we have looked into the ROI, how much this has improved the world. GDPR is consent-based privacy, but we have all been trained to accept all the cookie policies and so on [actually the ubiquity of being asked to accept cookies is not GDPR, it is the ePrivacy Directive]. You can put up any privacy policy you want, and users will still click accept. This type of policy even increases the digital divide. Those who are familiar with technology are more likely to know about GDPR, and know what rights they have and can exercise them. GDPR is just a good first step, but we need better regulations to improve privacy overall. I’m excited about where we will be in 5 years or so.

Q: Let's look at what has happened in the last few years with COVID. I know, Whitfield, that you've spent a lot of time thinking about this.

Whitfield: I think it is more than security. We used to have these discussions with retired military and academics under Chatham House rules [participants are free to use the information received, but neither the identity nor the affiliation of the speaker(s), nor that of any other participant, may be revealed]. Then we went online, and that had gone. There was no longer any informal privacy. On the other hand, we had people from Romania and from all over the world. I've also seen horror stories about exams where people were deemed to be cheating because they were gazing off into space when thinking about the answer.

Adi: There is also the difference between privacy and the perception of privacy. We created a privacy-preserving contact tracing app. But in spite of the security, the public was very suspicious and somehow felt privacy would be violated if they agreed to use the app. This led to the failure of contact tracing. But because of its quadratic nature, if only 10% download, the chance of finding all the contacts becomes very small. We had a good privacy-preserving app but could not convince the public.

Q: Let's wrap up with giving each of you a final statement.

Adi: The theme of this year's conference is "transform." If you look at the nature of warfare, received wisdom was that wars will be won and lost by cyber-attacks, but they have not happened as expected.

Whitfield: It seems to take the military a long time to learn how to use things.

Dawn. The key is blockchain so users control more and more of their cryptographic keys to control more and more types of assets. The next step after privacy-preserving and GDPR is to combine all these techniques.

Moni: I would like to see better dialogue between politicians and where definitions for things lie. Need better dialogue between regulators and the practitioners.

And with that Zulfikar thanked all the panelists.

 

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.

.