In this work we present two t-collusion resistant schemes to detect nodes performing 'Diversity Attacks' in network coding. This attack was introduced by Popa et al. in  and occurs when nodes performing network coding, do not code from some of their neighbors or they do not use random coefficients, which reduces network performance; this attack is fundamentally different from Pollution Attacks and it cannot be detected by current pollution prevention schemes. Our construction differs from the original solution in that it makes use of the broadcast nature of wireless networks; which allows us to reduce the computational load of both receivers and senders compared to their non-probabilistic solution; as an additional advantage, one of our constructions works without knowledge of the topology beyond the neighbors of a node, improving one of the shortcomings of the original solution. Tools employed in our constructions involve threshold digital signatures and aggregate Message Authentication Codes (MACs).