Abstract Deterministic Finite Automatons (DFAs) are widely used to perform regular expression based pattern matching for network security. Most of DFA compression algorithms consider only one type of redundant information in DFA. After a comprehensive analysis of different redundancies in DFA structure, we summarize four types of redundancies with large compression potentials. An improved DFA method, Tag-DFA, is proposed to exploit more than one kind of redundancy to compress DFA transitions and a reference table is added to accelerate DFA inspection. Compared with a well-known algorithm D2FA, Tag-DFA achieves more than 90% compression ratio and it requries at most two states traversal for each character, rather than multiple default transitions in D2FA. Despite a larger contruction complexity than D2FA, Tag-DFA guarantees a higher throughput with less memory cost.