This post is a response to a variety of discussions that I have had lately with a variety of customers, executives, salespeople, and even engineers that are working on security projects for a variety of companies. Sometimes, it seems, that encryption is positioned as the “Magic Bullet” that will cure all of their data protection woes. Worse yet, some will say that the “emperor has no clothes” or call their project “security theater.”
Encryption is not a magic bullet, but it does it play a vital role in a company’s data protection strategy. Most people are aware that there are two high-level types of encryption. These are encryption-at-rest and encryption-in-transit. Because an end-to-end encryption commitment uses both types of encryption in various ways, it is important to dive just a little deeper to eliminate any confusion when a cloud service provider is discussing encryption.
The security goal of encryption-at-rest is to provide access control by permitting only the authorized users or processes to view or alter the unencrypted data. As a result, it protects the integrity and confidentiality of the data. Unauthorized users or processes will see ciphertext. Ciphertext looks like random data and has no discernible patterns or meaning.
The most important thing to remember when thinking about encryption-at-rest is that the entity which holds the encryption/decryption key is responsible for enforcing access control. Encryption can be applied at different layers to mitigate various types of threats pertaining to unauthorized access. For example, sometimes this is implemented at the storage layer using self-encrypting disks. In this case, the storage operating system holds the key. If a disk drive is removed and connected to another system, the data will not be able to be read. It is important to note that if an attacker can come through the operating system that is authorized to use the storage system, this type of encryption will be ineffective. This kind of encryption is also known as transparent disk encryption because the encryption and decryption operations operate in a manner that is transparent to the operating system.
Moving up a layer, there is also transparent database encryption. The database system performs the encryption, and it is indeed transparent to the applications which store data in the database. Because the database management system holds the key, it has the access control responsibility. Access to the data must be made via the database management system, which will typically implement role-based access control. Anyone attempting to access the data by circumventing the database management system will only see ciphertext.
With application-layer encryption, the application holds the encryption key in memory and encrypts specific fields before inserting them into the database. Therefore, the application performs the access control duties. If the database administrator directly queries the database, all that will be returned are records that contain ciphertext.
So far we have talked about three different types of data encryption at rest, but what about the storage of the encryption keys? Typically, a Hardware Security Module (HSM) is used to store the keys and provides other important security features.
Encryption keys should never be stored with the application or the data they encrypt. Sadly, this is sometimes the case. Instead, when the application or database initializes, it will authenticate itself to the HSM and will obtain its keys using specific function calls that perform the key handling operations securely.
In addition to storing keys, Hardware Security Modules provide features for automating the key management processes. Any manual key handling processes require the cooperation of two different people to achieve what is known as “split control,” a security best practice. HSMs are also tamper-resistant and will shred the keys if someone tries to open the physical container.
The security goal of encryption-in-transit is to create a secure communication channel between two processes or people so as to protect the confidentiality and integrity of the information exchanged. The security of the channel is typically attacked by trying to eavesdrop or attempting to take a man-in-the-middle (MITM) position between the two parties that are communicating.
For an attacker to achieve a man-in-the-middle position, they must impersonate the client to the server and impersonate the server to the client. It is for this reason that it is imperative to authenticate the parties that are attempting to communicate securely. The server typically asserts its identity using a web server certificate that the client trusts. Sometimes the client will also use a certificate that the server trusts to authenticate itself to the server and other times the client authentication is left to the web application login functionality. Obviously, it is important that the login credentials are not passed to an imposter.
Many organizations are either moving to encryption between internal services or have already implemented it. By requiring mutual authentication and encryption for services that exchange sensitive data with other services within an organization’s production environment, this enables the “least privilege” security tenant. Each system or service only needs to talk to only a few other specific systems or services that need to consume its data.
What could go wrong?
Most people know better than to create their own cryptographic algorithms. Security through obscurity never works, especially with encryption. For this reason, it is important to use only ciphers that have withstood the test of time with lots of public scrutiny. Security should depend on the secrecy of the key and not the secrecy of the algorithm. Assuming that an industry accepted cipher such as the advanced encryption standard (AES) is used, there is still lots of room for implementation issues. Implementation issues could be another lengthy post, but two of the most common tend to be keeping both an unencrypted copy of the data along with the encrypted data or mishandling the encryption key. The main point here is that it is important to do a complete security analysis of an organization’s encryption rather than just to accept that they have the encryption magic bullet.