A project of the Centre for Internet and Society, India
Supported by Omidyar Network India
In an experiment to streamline its COVID-19 immunisation drive, India has adopted a centralised vaccine administration system called CoWIN (or COVID Vaccine Intelligence Network). In addition to facilitating registration for both online and walk-in vaccine appointments, the system also allows for the digital verification of vaccine certificates, which it issues to people who have received a dose. This development aligns with a global trend, as many countries have adopted or are in the process of adopting “vaccine passports” to facilitate safe movement of people while resuming commercial activity.
Some places, such as the EU, have constrained the scope of use of their vaccine certificates to international travel. The Indian government, however, has so far skirted important questions around where and when this technology should be used. By allowing anyone to use the online CoWIN portal to scan and verify certificates, and even providing a way for the private-sector to incorporate this functionality into their applications, the government has opened up the possibility of these digital certificates being used, and even mandated, for domestic everyday use such as going to a grocery shop, a crowded venue, or a workplace.
In this blog post, we examine the purported benefits of digital vaccine certificates over regular paper-based ones, analyse the privacy implications of their use, and present recommendations to make them more privacy respecting. We hope that such an analysis can help inform policy on appropriate use of this technology and improve its privacy properties in cases where its use is warranted.
We also note that while this post only examines the merits of a technological solution put out by the government, it is more important to consider the effects that placing restrictions on the movement of unvaccinated people has on their civil liberties in the face of a vaccine rollout that is inequitable along many lines, including gender, caste-class, and access to technology.
Every vaccine recipient in the country is required to be registered on the CoWIN platform using one of seven existing identity documents.1 Once a vaccine is administered, CoWIN generates a vaccine certificate which the recipient can access on the CoWIN website. The certificate is a single page document that contains the recipient’s personal information — their name, age, gender, identity document details, unique health ID, a reference ID — and some details about the vaccine given.2 It also includes a “secure QR code” and a link to CoWIN’s verification portal.
The verification portal allows for the verification of a certificate by scanning the attached QR code. Upon completion, the portal displays a success message along with some of the information printed on the certificate.
Verification is done using a cryptographic mechanism known as digital signatures, which are encoded into the QR code attached to a vaccine certificate. This mechanism allows “offline verification”, which means that the CoWIN verification portal or any private sector app attempting to verify a certificate does not need to contact the CoWIN servers to establish its authenticity. It instead uses a “public key” issued by CoWIN beforehand to verify the digital signature attached to the certificate.
The benefit of this convoluted design is that it protects user privacy. Performing verification offline and not contacting the CoWIN servers, precludes CoWIN from gleaning sensitive metadata about usage of the vaccine certificate. This means that CoWIN does not learn about where and when an individual uses their vaccine certificate, and who is verifying it. This closes off a potential avenue for mass surveillance.3 However, given how certificate revocation checks are being implemented (detailed in the privacy implications section below), CoWIN ends up learning this information anyway.
The primary argument for the adoption of digital verification of vaccine certificates over visual examination of regular paper-based ones is security. In the face of vaccine hesitancy, there are concerns that people may forge vaccine certificates to get around any restrictions that may be put in place on the movement of unvaccinated people. The use of digital signatures serves to allay these fears.
In its current form, however, digital verification of vaccine certificates is no more secure than visually inspecting paper-based ones. While the “secure QR code” attached to digital certificates can be used to verify the authenticity of the certificate itself, the CoWIN verification portal does not provide any mechanism nor does it instruct verifiers to authenticate the identity of the person presenting the certificate. This means that unless an accompanying identity document is also checked, an individual can simply present someone else’s certificate.
There are no simple solutions to this limitation; adding a requirement to inspect identity documents in addition to digital verification of the vaccine certificate would not be a strong enough security measure to prevent the use of duplicate vaccine certificates. People who are motivated enough to forge a vaccine certificate, can also duplicate one of the seven ID documents which can be used to register on CoWIN, some of which are simple paper-based documents.4 Requiring even stronger identity checks, such as the use of Aadhaar-based biometrics, would make digital verification of vaccine certificates more secure. However, this would be a wildly disproportionate incursion on user privacy — allowing for the mass collection of metadata like when and where a certificate is used — something that digital vaccine certificates were explicitly designed to prevent. Additionally, in Russia, people were found issuing fake certificates by discarding real vaccine doses instead of administering them. No technological solution can prevent such fraud.
As such, the utility of digital certificates is limited to uses such as international travel, where border control agencies already have strong identity checks in place for travellers. Any everyday usage of the digital verification functionality on vaccine certificates would not present any benefit over visually examining a piece of paper or a screen.
In addition to providing little security utility over manual inspection of certificates, digital certificates also present privacy issues, these are listed below along with recommendations to mitigate them:
(i) The verification portal leaks sensitive metadata to CoWIN’s servers: An analysis of network requests made by the CoWin verification portal reveals that it conducts a ‘revocation check’ each time a certificate is verified. This check was also found in the source code, which is made openly available.5
Revocation checks are an important security consideration while using digital signatures. They allow the issuing authority (CoWIN, in this case) to revoke a certificate in case the account associated with it is lost or stolen, or if a certificate requires correction. However, the way they have been implemented here presents a significant privacy issue. Sending certificate details to the server on every verification attempt allows it to learn about where and when an individual is using their vaccine certificate.
We note that the revocation check performed by the CoWIN portal does not necessarily mean that it is storing this information. Nevertheless, sending certificate information to the server directly contradicts claims of an “offline verification” process, which is the basis of the design of these digital certificates.
Recommendations: Implementing privacy-respecting revocation checks such as Certificate Revocation Lists,6 or Range Queries7 would mitigate this issue. However, these solutions are either complex or present bandwidth and storage tradeoffs for the verifier.
(ii) Oversharing of personally identifiable information: CoWIN’s vaccine certificates include more personally identifiable information (name, age, gender, identity document details and unique health ID) than is required for the purpose of verifying the certificate. An examination of the vaccine certificates available to us revealed that while the Aadhaar number is appropriately masked, other personal identifiers such as passport number and unique health ID were not masked. Additionally, the inclusion of demographic details, such as age and gender, provides little security benefit by limiting the pool of duplicate certificates that can be used and are not required in light of the security analysis above.
Recommendation: Personal identifiers (such as passport number and unique health ID) should be appropriately masked and demographic details (age, gender) can be removed.
The minimal set of data required for identity-linked usage for digital verification, as described above, is a full name and masked ID document details. All other personally identifying information can be removed. In case of paper-based certificates, which is suggested for domestic usage, only the details about vaccine validity would suffice and no personal information is required.
(iii) Making information available digitally increases the likelihood of collection: All of the personal information printed on the certificate is also encoded into the QR code. This is necessary because the digital signature verification process also verifies the integrity of this information (i.e. it wasn’t modified). A side effect of this is that the personal information is made readily available in digital form to verifiers when it is scanned, making it easy for them to store. This is especially likely in private sector apps who may be interested in collecting demographic information and personal identifiers to track customer behaviour.
Recommendation: Removing extraneous information from the certificate, as suggested above, mitigates this risk as well.
Our analysis reveals that without incorporating strong, privacy-invasive identity checks, digital verification of vaccine certificates does not provide any security benefit over manually inspecting a piece of paper. The utility of digital verification is limited to purposes that already conduct strong identity checks.
In addition to their limited applicability, in their current form, these digital certificates also generate a trail of data and metadata, giving both government and industry an opportunity to infringe upon the privacy of the individuals using them.
Keeping this in mind, the adoption of this technology should be discouraged for everyday use.
|1||Exceptions exist for people without state-issued identity documents. ↑|
|2||This information was gathered by inspecting three vaccine certificates linked to the author’s CoWIN account, which they were authorised to view, and may not be fully accurate. ↑|
|3||This design is similar to Aadhaar’s “offline KYC” process. ↑|
|4||“Aadhaar Card: UIDAI says downloaded versions on ordinary paper, mAadhaar perfectly valid”, Zee Business, April 29 2019, https://www.zeebiz.com/india/news-aadhaar-card-uidai-says-downloaded-versions-on-ordinary-paper-maadhaar-perfectly-valid-96790. ↑|
|5||This check was also verified to be present in the reference code made available for private-sector applications incorporating this functionality, suggesting that private sector apps will also be affected by this. ↑|
|6||Certificate Revocation Lists allow the server to provide a list of revoked certificates to the verifier, instead of the verifier querying the server each time. This, however, can place heavy bandwidth and storage requirements on the verifying app as this list can potentially grow long. ↑|
|7||Range Queries are described in this paper. In this method, the verifier requests revocation status from the server by specifying a range of certificate identifiers within which the certificate being verified lies. If there are any revoked certificates within this range, the server will send their identifiers to the verifier, who can then check if the certificate in question is on the list. For this to work, the range selected must be sufficiently large to include enough potential candidates to keep the server from guessing which one is in use. ↑|