COMMON WEAKNESS ENUMERATIONIncomplete

CWE-1039

Inadequate Detection or Handling of Adversarial Input Perturbations in Automated Recognition Mechanism

The product uses an automated mechanism such as machine learning to recognize complex data inputs (e.g. image or audio) as a particular concept or category, but it does not properly detect or handle inputs that have been modified or constructed in a way that causes the mechanism to detect a different, incorrect concept.

Abstraction

Class

Structure

Simple

Likelihood of Exploit

Medium

Understand the problem

What is the problem?

The product uses an automated mechanism such as machine learning to recognize complex data inputs (e.g. image or audio) as a particular concept or category, but it does not properly detect or handle inputs that have been modified or constructed in a way that causes the mechanism to detect a different, incorrect concept.

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.

Root cause

How and when does the problem arise?

Lifecycle phase

Architecture and Design

This issue can be introduced into the automated algorithm itself due to inadequate training data used as well as lack of validation, verification, testing, and evaluation of the algorithm. These factors can affect the overall robustness of the algorithm when introduced into operational settings.

Lifecycle phase

Implementation

The developer might not apply external validation of inputs into the algorithm.

Risk

What does successful exploitation lead to?

Bypass Protection Mechanism

Affected security scopes: Integrity

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.
DoS: Resource Consumption (Other)DoS: Instability

Affected security scopes: Availability

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

Affected security scopes: Confidentiality

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.
Varies by Context

Affected security scopes: Other

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

Defence

How is it prevented and fixed?

Architecture and Design

Architecture and design approach

Detail
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.

Architecture and Design

Architecture and design approach

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

Architecture and Design

Architecture and design approach

Detail
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.

Implementation

Implementation-phase defence

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

Implementation

Implementation-phase defence

Detail
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.

Integration

Integration aşaması savunması

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

Integration

Integration aşaması savunması

Detail
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.

Verification

How is it detected?

Dynamic Analysis with Manual Results Interpretation

Use indicators from model performance deviations such as sudden drops in accuracy or unexpected outputs to verify the model.

Dynamic Analysis with Manual Results Interpretation

Use indicators from input data collection mechanisms to verify that inputs are statistically within the distribution of the training and test data.

Architecture or Design Review

Use multiple models or model ensembling techniques to check for consistency of predictions/inferences.