On Friday of the first week of classes, two hours before the registration deadline, I signed up to take a privacy-oriented class with Dr. Sauvik Das. I hoped to learn more about privacy and security, and my first day didn't disappoint. I heard the phrase "homomorphic encryption" mentioned in an offhand comment, and my curiosity led me to learn about a very interesting problem I had never heard of before!
Before 2009, there was a open-ended problem that plagued the exploding field of cloud-computing. Say I was a privacy-concerned individual and I uploaded thousands and thousands of encrypted, private banking documents to Google servers. Then, I asked that the servers apply a 10% tax on all of my account balances. There was no way to complete this operation without individually decrypting every single document, finding the plaintext values, applying the 10% reduction, and re-encrypting it. Most people didn't trust cloud services with the keys to decrypt their documents in the first place, and thus large-scale external servers generally weren't used to store and operate on private data. This limitation was an obstacle to many awesome applications of cloud computing, such as large-scale processing of protected patient health data and anonymous citizen voting data.
The term 'homomorphic encryption' refers to an encryption scheme that allows operations to be applied to encrypted data without ever decrypting it . For example, say I upload an encrypted message to Gmail servers to my friend that says "hey man, what's up?". I decide that the message doesn't seem enthusiastic enough, so I ask Google to capitalize every letter such that the new message is "HEY MAN, WHAT'S UP?". If I used a homomorphic encryption scheme for my message, Google can make that change without ever knowing what's in the message. This was purely conceptual for quite a long time, and many scientists pursued a solution for around 30 years. A number of 'semi-homomorphic' encryption schemes arose, and you can guess, these schemes could perform some operations such as addition on an encrypted number, but not arbitrary operations such as capitalization.
In an extremely impressive Ph.D dissertation by Dr. Gentry in 2009, the outline for a fully homomorphic encryption scheme was published. The specifics of the solution are a bit math-heavy and hard for myself to understand, but the underlying idea was to combine two semi-homomorphic schemes into a much stronger one. By cleverly combining these existing solutions, he devised a system in which a NAND operation could be performed on the bits of a ciphertext while maintaining the original encryption scheme. If you don't have an introduction to computation, just about any operation you could dream of is implementable by some combination of NAND operations. By creating this scheme, he solved the much sought-after problem of fully homomorphic encryption (FHE).
Dr. Gentry is now an industry researcher working at IBM's Watson Research Center, and has obtained a law degree from Harvard. He has gone on to improve his FHE designs and produce an insane amount of contributions to the cryptographic community.
One of the most recent health reports released by the CDC in November 2018 was a presentation and statistical analysis of drug overdose deaths for the most part of the 21st Century. For better or for worse these reports don't contain any analysis into the social or environmental causes of health trends, and although this is perhaps the best way to remain unbiased it does leave quite a lot of questions unanswered.
Since 1999, the rate of drug overdose deaths has been steadily rising up and up, with men aged 25-54 being significantly more prevalent than other age/gender groups. In 2015 there was a huge spike that saw the rate for this particular demographic surge up almost 25%, and this is the largest surge that we've seen in at least the past 20 years or so. What could be the cause of this? Opioids such as heroin and methadone have always been some of the worst culprits for lethal overdose, but since 2012 the death rate from synthetic opioids has increased more then ten-fold. One such opioid, fentanyl, seems to be at the forefront of this synthetic drug boom that has been responsible for recent deaths such as celebrities Mac Miller, Prince, and Tom Petty. After doing some digging around, it seems like fentanyl roughly is to heroin as spice is to marijuana.
Marijuana and heroin aren't two drugs that have a lot in common, but they're both derived from natural plants: hemp and poppy. Unless these drugs have been cut with another substance, their chemical makeup is generally uniform and their production is limited to the growth of their respective plant source. Spice is a synthetic imitation of marijuana produced in a lab at scale, and has become particularly popular in the UK. To avoid specific legislation aimed at its natural alternative, producers of spice alter the chemical composition of their product however they like so that it's not recognized as marijuana under current drug laws, and thus distributers can avoid repurcussions even if they are caught. Since spice can be made non-uniformly and at-scale, it is almost infinitely more dangerous than marijuana. In a similar way, fentanyl is a synthetic imitation of heroin. It's far more potent, cheaper to produce, and easier to make more of. Producers and distributers have noticed this, and have begun to make changes in their market accordingly. Fentanyl is also recognized as a presciption drug in the US with a not-insignificant population of legal users for pain medication, and so it's harder for drug agencies to aggressively pursue enforcing consequences for carrying fentanyl.
This is clearly an epidemic that is becoming a bigger issue day by day; what needs to happen from here? The reasoning I've read in some articles suggests that a number of overdoses are a consequence of the user's misinformation. Heroin has been in our society for many years, and addict communities are generally knowledged enough to know what a lethal overdose is. Many of these people are now being handed fentanyl, and treat the new drug as if it were heroin in terms of dosage. Doing so will drastically increase the chances of death, as fentanyl is significantly more potent and should have a much lower dosage. Taking a liberal approach and conducting an information campaign to educate communities about the properties of fentanyl could go a long way in reducing deaths. I'm personally not knowledged enough about the subject, but if there are clear distinctions between the physical properties of heroin and fentanyl then spreading this information could be valuable as well.
Looking at visual representations of the data included in this CDC report is pretty depressing, and unless something is done soon it's likely this spike in drug deaths due to this chemical will continue to skyrocket.