Stronghold Threat Modeling
As a security software, Stronghold prevents attackers from accessing the secrets that it stores. Threat modeling is a systematic approach that tries to assess all the potential attack scenarios on software.
There are multiple steps to this process:
- Model your software.
- Define the assets and security property you protect.
- List potential threats using existing classification (STRIDE in our case).
- Propose mitigations to those threats
Stronghold’s Model
Stronghold is simple in concept; it is used to store secrets. These secrets should never be revealed, even to their owners. Users can interact with secrets through controlled methods called procedures.
Typical use of Stronghold
Typical use of Stronghold
- Generate a key in Stronghold.
- Use Stronghold procedures to use the key for encryption, decryption, signatures, etc.
- Store Stronghold state for future usage in permanent storage called Snapshot.
Model
- Users can only interact with secrets through the procedures API.
- Users can only use procedures on the secrets they own
- Procedures cannot reveal/temper secrets (can delete them though)
- Secrets can be kept permanently in an encrypted form in a Snapshot (filesystem)
Assets
The sole asset of Stronghold is its secrets.
High Importance
- Confidentiality: Secrets are never revealed.
- IntegrityAssurance that the data stored in the ledger has not been tampered with or altered in any unauthorized way.: Secrets cannot be modified, only deleted.
- Authentication: Only authorized users can interact with secrets,
Mid Importance
- Availability: A user can interact with its secrets anytime.
Low Importance
- Least privilege: There are no privileged users in Stronghold.
- Non-repudiation: A user can't disprove that it has used a procedure on a secret.
Attack Surface
Stronghold defends against multiple types of attackers. Levels also represent how likely it is for an attacker to appear.
- Level 1: Procedure API.
- Level 2: Permanent storage: Snapshot in the filesystem.
- Level 3: Memory:
- The attacker can read memory.
- Through cold-boot attacks or memory dumps.
- Level 4: Side-channels:
- Timing attacks.
- Power consumption.
A type of attacker that we don't represent here but is also important comes from potential vulnerabilities in the tools used to build the software.
- Bonus: Tools:
- Packages.
- Rust language.
- Compilation.
- Crypto algorithms used.
Potential threats STRIDE
We use the STRIDE threats classification. Stride is applied to all the types of attacks listed above. Mentions of WIP means that it is still "Work In Progress".
Level 1: Procedure API
Attack | Attack | Remediation | Severity |
---|---|---|---|
Spoofed | An unauthorized user executes procedures | When restoring Stronghold state from a snapshot a key is required to decrypt the snapshot | High |
Tampered | Secrets are tampered with using procedures | Procedures are developed and audited by the team so they don't modify secrets | High |
Repudiated | A user is accused to have used secrets maliciously/incorrectly | Log all the procedures that have been processed (WIP) | Low |
Information Disclosure | A secret is revealed through procedures | Procedures are developed and audited by the team to ensure they don't reveal secrets | High |
Denial of Service | Spamming procedures to block the system | This is the responsability of the software that uses | Mid |
Elevation of Privileges | None, there are no privileged users in Stronghold |
Level 2: Permanent Storage, File System
Attack | Attack | Remediation | Severity |
---|---|---|---|
Spoofed | OS responsibility | ||
Tampered | Files storing the secrets are modified. Secrets can be lost. | Check the integrity of snapshot with a checksum. Keep an older snapshot to be able to restore the correct state | High |
Repudiated | OS responsibility | ||
Information Disclosure | Snapshot content is read | Snapshot content is encrypted | High |
Denial of Service | Host file system is unavailable. Stronghold cannot commit its current state or load a previous state | Stronghold can continue, but can't commit | Mid |
Elevation of Privileges | Attacker has elevated privilege on the host machine, and can read, write or delete snapshots on the file system | Same case as Tampered and Info Disclosure attacks, the snapshot is encrypted and can be restored if it has a valid checksum | High |
Level 3: Memory
Attack | Attack | Remediation | Severity |
---|---|---|---|
Spoofed | OS responsibility | ||
Tampered | Host system gets its memory corrupted. Procedures will produce wrong outputs, so original data can be lost | Secrets are backed by permanent a storage called Snapshot. User may use a previous snapshot to restore a previous state | High |
Repudiated | OS responsibility | ||
Information Disclosure | Secrets are revealed through reading the memory directly | Secrets are stored encrypted in the memory and are only decrypted for the minimum amount of time. Moreover the Boojum scheme protects encryption keys in memory. | High |
Denial of Service | Memory is not accessible, preventing Stronghold from working | OS responsibility | Mid |
Elevation of Privileges | Attacker has elevated privileges on the host machine and can access the secrets in Stronghold | Same case as Tampered and Info Disclosure attacks, the snapshot is encrypted and can be restored if it has a valid checksum | High |
Level 4: Side-channels
Attack | Attack | Remediation | Severity |
---|---|---|---|
Spoofed | No potential attack | ||
Tampered | Tamper memory using side-channels | Refer to memory and storage tables | High |
Repudiated | No potential attack | ||
Information Disclosure | Secrets are revealed through side-channels | Make sure the procedure are constant in time and energy usage (WIP). This also depends on the cryptographic implementation used | High |
Denial of Service | Prevent normal behaviour of stronghold through side-channels such as electromagnetic waves | You can't protect this from software, it is the host's responsibility | Mid |
No potential attack |