Is homomorphic encryption the Holy Grail of data security?

In ‘locked-room’ mys­ter­ies, a crime has been com­mit­ted in a room that no one could have entered, and the read­er has to work out what hap­pened. Spoil­er: the room was nev­er as secure as it seemed. And real-life data crime has, unfor­tu­nate­ly, the same spoil­er. Unen­crypt­ed data is nev­er real­ly secure – no mat­ter how good the locks on the perime­ter may look. 

The obvi­ous solu­tion to that, of course, is to keep data encrypt­ed. But what hap­pens when you need to use it? Until recent­ly, the need to have data in plain text to be able to analyse it was an insur­mount­able secu­ri­ty prob­lem – par­tic­u­lar­ly on the cloud and when data was being shared. Now, how­ev­er, firms say there is a way to have secure data shar­ing and col­lab­o­ra­tion: homo­mor­phic encryp­tion.

A leap of faith? 

Homo­mor­phic encryp­tion has long been the holy grail of cyber­se­cu­ri­ty. IBM’s cryp­tog­ra­phy expert Craig Gen­try defined it as: “[how] a third par­ty can per­form com­pli­cat­ed pro­cess­ing of data with­out being able to see it”. The anal­o­gy Gen­try gives is of a jew­ellery work­shop with a locked box of pre­cious mate­ri­als that only the own­er can open. Staff, how­ev­er, can access the box using gloves to assem­ble the jew­ellery but can’t take any­thing out, leav­ing the assem­bled jew­ellery safe for the own­er.

Homo­mor­phic encryp­tion is sim­i­lar in that it lets data proces­sors manip­u­late select­ed ‘raw mate­ri­als’ such as sales fig­ures or med­ical data, but keeps the plain-text data pri­vate. That is because the data doesn’t have to be decrypt­ed to be used. Only the end result of the com­pu­ta­tion is pre­sent­ed in plain text. Because homo­mor­phic encryp­tion is math­e­mat­i­cal­ly and com­pu­ta­tion­al­ly very chal­leng­ing, it was an intrigu­ing the­o­ret­i­cal dis­cus­sion long before it became a prac­ti­cal option. And it is still in devel­op­ment.

“There are no the­o­ret­i­cal lim­its to the com­pu­ta­tions that can be car­ried out using homo­mor­phic encryp­tion,” says Elli­son Anne Williams, CEO and founder of Enveil, a com­pa­ny that spe­cialis­es in pri­va­cy enhanc­ing tech­nolo­gies. “But there are prac­ti­cal con­straints.” In par­tic­u­lar, homo­mor­phic encryp­tion is still lim­it­ed in terms of the func­tions it can car­ry out and it needs a lot of pro­cess­ing pow­er to run. 

Why use homomorphic encryption?

Giv­en that analysing data you can’t see seems to require both a leap of faith (at least for non-math­e­mati­cians) and sig­nif­i­cant resources, why use it? Williams says that, at Enveil, they “don’t ask clients to press the ‘I believe’ but­ton” but pro­vide frame­works and tools to ver­i­fy the analy­sis and to help peo­ple under­stand what is going on. 

The use cas­es are also increas­ing­ly evi­dent. Robert Schukai, exec­u­tive vice-pres­i­dent, tech­nol­o­gy devel­op­ment, fin­tech and new infra­struc­ture at Mas­ter­card (an investor in Enveil), said in the keynote address at the 2021  Secure and Pri­vate Com­pute Sum­mit: “Homo­mor­phic encryp­tion is a phe­nom­e­nal­ly excit­ing tech­nol­o­gy. We see great val­ue in query­ing data where it lives… Homo­mor­phic encryp­tion is an ide­al tech­nol­o­gy when you are deal­ing with sen­si­tive data that you don’t want to sling around but would pre­fer to leave in its loca­tion.”

Firms often need to inter­ro­gate data that is kept else­where. Big multi­na­tion­als, for exam­ple, share infor­ma­tion across bor­ders. Homo­mor­phic encryp­tion allows them to do that while still meet­ing local data and reg­u­la­to­ry require­ments because what is moved around is the analy­sis of the data. The data itself is kept in place and remains encrypt­ed. 

Homo­mor­phic encryp­tion is already in use by the large firms that can afford to pay for, and val­ue, the use cas­es. Williams points out, for exam­ple, that ran­somware goes after data at rest, so firms are well advised to keep data per­ma­nent­ly encrypt­ed if they can.  

On the horizon for clouds

Homo­mor­phic encryp­tion promis­es that data nev­er has to be moved or pre­sent­ed in plain text. Even if there is a perime­ter breach, the data is safe. But is homo­mor­phic encryp­tion itself a secure­ly locked room? What about quan­tum com­put­ing, which is pow­er­ful enough to break many of the ciphers now in use? Williams says that even quan­tum com­put­ing won’t be able to crack homo­mor­phic encryp­tion because it doesn’t rely on fac­tor­ing huge num­bers. 

Unsur­pris­ing­ly, giv­en the promis­es of the tech­nol­o­gy, the big cloud play­ers are all active in homo­mor­phic encryp­tion. Microsoft, for exam­ple, offers Microsoft SEAL (which stands for sim­ple encrypt­ed arith­metic library) and does what it says on the tin. It is a set of ‘encryp­tion libraries’ that helps soft­ware engi­neers build end-to-end encrypt­ed ser­vices. The open-source tech­nol­o­gy aims to make homo­mor­phic encryp­tion ‘easy to use and avail­able to every­one’ – not just peo­ple with a deep under­stand­ing of the com­plex maths.

Google launched its open-source ful­ly homo­mor­phic encryp­tion library in June this year. Again, the aim is to bring every­one on board with open-source soft­ware. Google’s solu­tion is a tran­spiler that turns code for “any type of basic com­pu­ta­tion… into a ver­sion that can run on encrypt­ed data”. 

Miguel Gue­vara, prod­uct man­ag­er in Google’s pri­va­cy team, says: “Up to our release, you need­ed to be expert to pro­duce things on top of encrypt­ed data. This tool lets any devel­op­er do it. You no longer need a PhD in the field.” How­ev­er, there may still be a big gap between a library that researchers and devel­op­ers use and a solu­tion that busi­ness­es can imple­ment. This isn’t one to try at home.

Also, Gue­vara says that while Google’s offer is “very good for basic things such as ver­i­fy­ing an age in a data­base, or updat­ing and chang­ing records… we’re far away from being able to con­vert all appli­ca­tions to ful­ly homo­mor­phic encryp­tion”. It’s also still cloud-based and not for ‘edge devices’ like mobile phones.

“That is most­ly because the tech­nol­o­gy is still very new,” he says. “Over time, there will be a mix. For exam­ple, homo­mor­phic encryp­tion could be used to hold the keys to the data on a phone.”

Still, even now, homo­mor­phic encryp­tion promis­es to solve some major pri­va­cy prob­lems as well as eas­ing secu­ri­ty headaches. For exam­ple, Gue­vara says that a data­base of smart devices in a home could be inter­ro­gat­ed to pro­duce a video snip­pet that pro­vides infor­ma­tion on the struc­ture of the home with­out expos­ing videos of the home. 

If firms – and gov­ern­ments – real­ly only do have access to anonymised data that is essen­tial to a par­tic­u­lar, and nec­es­sary, query, it won’t just be big, cloud-based com­pa­nies that get com­fort from homo­mor­phic encryp­tion.