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:

1. Model your software.
2. Define the assets and security property you protect.
3. List potential threats using existing classification (STRIDE in our case).
4. 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

1. Generate a key in Stronghold.
2. Use Stronghold procedures to use the key for encryption, decryption, signatures, etc.
3. 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.
• Integrity: 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
Elevation of Privileges	No potential attack