SERENE: A Collusion Resilient Replication-based Verification Framework

TL;DR

SERENE tackles collusion in replication-based task verification for edge computing in autonomous driving. It introduces a lightweight, single-task based detection and a three-stage mitigation that partitions workers and isolates colluders without relying on trusted servers. Empirical results show up to 60% gains in mitigation accuracy and up to 50% faster detection over the Staab & Angel baseline, with strong robustness even when colluders form a majority. The approach also demonstrates practical feasibility with modest resource usage on common edge devices, indicating real-world applicability for secure remote computation in autonomous systems.

Abstract

The rapid advancement of autonomous driving technology is accompanied by substantial challenges, particularly the reliance on remote task execution without ensuring a reliable and accurate returned results. This reliance on external compute servers, which may be malicious or rogue, represents a major security threat. While researchers have been exploring verifiable computing, and replication-based task verification as a simple, fast, and dependable method to assess the correctness of results. However, colluding malicious workers can easily defeat this method. Existing collusion detection and mitigation solutions often require the use of a trusted third party server or verified tasks which may be hard to guarantee, or solutions that assume the presence of a minority of colluding servers. We propose SERENE, a collusion resilient replication-based verification framework that detects, and mitigates colluding workers. Unlike state-of-the-art solutions, SERENE uses a lightweight detection algorithm that detects collusion based on a single verification task. Mitigation requires a two stage process to group the workers and identifying colluding from honest workers. We implement and compare SERENE's performance to Staab et. al, resulting in an average of 50\% and 60\% accuracy improvement in detection and mitigation accuracy respectively.

Figures (7)

• Figure 1: In time t, there is just a group of majority ($V^i=R$), but in $t+\Delta t$, the received result $R'$ makes the second majority group, and collusion detected.
• Figure 2: Group identification example: In Case (I), where G1 consists of honest workers, SERENE selects pools P' entirely from G2; nodes disagreeing with majority are honest and added to G1; In Case (II), where G1 consists of colluding workers, P' is formed with two nodes from G1 and one from G2, until we verify all G2 members ( i.e., G2 members agreeing with majority are colluding and added to G1).
• Figure 3: Comparing SERENE's and SnE's collusion detection delay (a) and (b) and accuracy (c); INF denotes infinite delay values due to unsuccessful collusion detection
• Figure 4: Collusion detection delay as a function of task repository size, $L$
• Figure 5: Comparing mitigation accuracy (a, b) and latency (c) for SnE, SERENE implementation up to the grouping phase ( SERENE-Prt), SERENE implementation up to the identification of G1 ( SERENE-Prt+G1), and SERENE ( i.e., the full design).