[Paper Review] Susceptibility of Age of Gossip to Timestomping

Priyanka Kaswan, Şennur Ulukuş|arXiv (Cornell University)|May 17, 2022

Age of Information Optimization4 citations

TL;DR

This paper studies the impact of timestomping attacks on information freshness in fully connected gossip networks, showing that a single compromised node can degrade end-to-end age from O(log n) to O(n). The adversary optimally manipulates timestamps to make stale packets appear fresh and blocks incoming fresh packets, exploiting the network's full connectivity to maximize staleness.

ABSTRACT

We consider a fully connected network consisting of a source that maintains the current version of a file, $n$ nodes that use asynchronous gossip mechanisms to disseminate fresh information in the network, and an adversary who infects the packets at a target node through data timestamp manipulation, with the intent to replace circulation of fresh packets with outdated packets in the network. We show that a single infected node increases the expected age of a fully connected network from $O(\log n)$ to $O(n)$. Further, we show that the optimal behavior for an adversary is to reset the timestamps of all outgoing packets to the current time and of all incoming packets to an outdated time. Additionally, if the adversary allows the infected node to accept a small fraction of incoming packets from the network, then a large network can manage to curb the spread of stale files coming from the infected node and pull the network age back to $O(\log n)$. Lastly, we show that if an infected node contacts only a single node instead of all nodes of the network, the system age can still be degraded to $O(n)$. These show that fully connected nature of a network can be both a benefit and a detriment for information freshness; full connectivity, while enabling fast dissemination of information, also enables fast dissipation of adversarial inputs.

Motivation & Objective

• To analyze the susceptibility of age-of-gossip in fully connected networks to timestomping attacks by an adversary.
• To determine the optimal adversarial strategy for maximizing network age through timestamp manipulation.
• To evaluate the resilience of gossip networks when infected nodes accept a small fraction of fresh packets.
• To assess whether limited connectivity (e.g., contacting only one node) still enables significant age degradation.

Proposed method

• Models a fully connected network with a source, n gossiping nodes, and one adversarially compromised node.
• Uses stochastic hybrid systems to derive recursive age equations for node states under timestomping.
• Analyzes age dynamics by partitioning node states based on the number of neighbors that have received fresh data.
• Applies probabilistic bounds and recursive inequalities to derive lower bounds on expected age at regular nodes.
• Considers three attack models: full timestomping (outgoing to current time, incoming to outdated time), partial acceptance (p=0.99), and isolated node scenarios.
• Employs numerical simulations with λ=1 and total time 1000n to validate analytical results on age scaling.

Experimental results

Research questions

• RQ1 How does timestomping by a single compromised node affect the expected age of information in a fully connected gossip network?
• RQ2 What is the optimal timestamp manipulation strategy for an adversary to maximize network age?
• RQ3 Can a network recover freshness if the infected node accepts even a small fraction of incoming fresh packets?
• RQ4 Does the age degradation persist when the infected node communicates with only one other node instead of all nodes?

Key findings

• A single infected node increases the expected age of all regular nodes from O(log n) to O(n) due to timestomping.
• The optimal adversarial strategy is to set all outgoing packet timestamps to the current time and all incoming packet timestamps to an outdated time.
• If the infected node accepts even 1% of incoming fresh packets (p=0.99), the network age scales back to O(log n) for large n.
• Even when the infected node contacts only one other node, the system age degrades to O(n), showing that full connectivity is not required for severe impact.
• The age at a regular node is bounded below by v_A/4, where v_A is the age at the adversary node, which scales as O(n).
• Numerical results confirm analytical predictions: age at the infected node grows linearly with n when p=1, and logarithmically when p is close to 1.