CWE-1039

Extended description

When techniques such as machine learning are used to automatically classify input streams, and those classifications are used for security-critical decisions, then any mistake in classification can introduce a vulnerability that allows attackers to cause the product to make the wrong security decision or disrupt service of the automated mechanism. If the mechanism is not developed or "trained" with enough input data or has not adequately undergone test and evaluation, then attackers may be able to craft malicious inputs that intentionally trigger the incorrect classification. Targeted technologies include, but are not necessarily limited to: automated speech recognition automated image recognition automated cyber defense Chatbot, LLMs, generative AI For example, an attacker might modify road signs or road surface markings to trick autonomous vehicles into misreading the sign/marking and performing a dangerous action. Another example includes an attacker that crafts highly specific and complex prompts to "jailbreak" a chatbot to bypass safety or privacy mechanisms, better known as prompt injection attacks.

Common consequences4

• IntegrityBypass Protection Mechanism

  When the automated recognition is used in a protection mechanism, an attacker may be able to craft inputs that are misinterpreted in a way that grants excess privileges.

• AvailabilityDoS: Resource Consumption (Other)DoS: Instability

  There could be disruption to the service of the automated recognition system, which could cause further downstream failures of the software.

• ConfidentialityRead Application Data

  This weakness could lead to breaches of data privacy through exposing features of the training data, e.g., by using membership inference attacks or prompt injection attacks.

• OtherVaries by Context

  The consequences depend on how the application applies or integrates the affected algorithm.

Potential mitigations7

1. Architecture and Design

   Algorithmic modifications such as model pruning or compression can help mitigate this weakness. Model pruning ensures that only weights that are most relevant to the task are used in the inference of incoming data and has shown resilience to adversarial perturbed data.

2. Architecture and Design

   Consider implementing adversarial training, a method that introduces adversarial examples into the training data to promote robustness of algorithm at inference time.

3. Architecture and Design

   Consider implementing model hardening to fortify the internal structure of the algorithm, including techniques such as regularization and optimization to desensitize algorithms to minor input perturbations and/or changes.

4. Implementation

   Consider implementing multiple models or using model ensembling techniques to improve robustness of individual model weaknesses against adversarial input perturbations.

5. Implementation

   Incorporate uncertainty estimations into the algorithm that trigger human intervention or secondary/fallback software when reached. This could be when inference predictions and confidence scores are abnormally high/low comparative to expected model performance.

6. Integration

   Reactive defenses such as input sanitization, defensive distillation, and input transformations can all be implemented before input data reaches the algorithm for inference.

7. Integration

   Consider reducing the output granularity of the inference/prediction such that attackers cannot gain additional information due to leakage in order to craft adversarially perturbed data.

Relationships2

• ChildOfCWE-693
• ChildOfCWE-697