vLLM: Quadratic Time Complexity in Input Token Processing​ leads to denial of service

Summary

A critical performance vulnerability has been identified in the input preprocessing logic of the multimodal tokenizer. The code dynamically replaces placeholder tokens (e.g., <|audio_|>, <|image_|>) with repeated tokens based on precomputed lengths. Due to ​​inefficient list concatenation operations​​, the algorithm exhibits ​​quadratic time complexity (O(n²))​​, allowing malicious actors to trigger resource exhaustion via specially crafted inputs.

Details

​​Affected Component​​: input_processor_for_phi4mm function. https://github.com/vllm-project/vllm/blob/8cac35ba435906fb7eb07e44fe1a8c26e8744f4e/vllm/model_executor/models/phi4mm.py#L1182-L1197

The code modifies the input_ids list in-place using input_ids = input_ids[:i] + tokens + input_ids[i+1:]. Each concatenation operation copies the entire list, leading to O(n) operations per replacement. For k placeholders expanding to m tokens, total time becomes O(kmn), approximating O(n²) in worst-case scenarios.

PoC

Test data demonstrates exponential time growth:

test_cases = [100, 200, 400, 800, 1600, 3200, 6400]
run_times = [0.002, 0.007, 0.028, 0.136, 0.616, 2.707, 11.854]  # seconds

Doubling input size increases runtime by ~4x (consistent with O(n²)).

Impact

​​Denial-of-Service (DoS):​​ An attacker could submit inputs with many placeholders (e.g., 10,000 <|audio_1|> tokens), causing CPU/memory exhaustion. Example: 10,000 placeholders → ~100 million operations.

Remediation Recommendations​

Precompute all placeholder positions and expansion lengths upfront. Replace dynamic list concatenation with a single preallocated array.

# Pseudocode for O(n) solution
new_input_ids = []
for token in input_ids:
    if token is placeholder:
        new_input_ids.extend([token] * precomputed_length)
    else:
        new_input_ids.append(token)