The parametrization of (3, 3)-isogenies by Bruin, Flynn and Testa requires over 37.500 multiplications if one wants to evaluate a single isogeny in a point. We simplify their formulae and reduce the amount of required multiplications by 94%. Further we deduce explicit formulae for evaluating (3, 3)-splitting and gluing maps in the framework of the parametrization by Bröker, Howe, Lauter and Stevenhagen. We provide implementations to compute (3 n , 3 n )-isogenies between principally polarized abelian surfaces with a focus on cryptographic application. Our implementation can retrieve Alice's secret isogeny in 11 seconds for the SIKEp751 parameters, which were aimed at NIST level 5 security.standardization process for public key exchanges [23]. That same month however, Castryck and Decru published a devastating attack on SIKE, retrieving Bob's private key in minutes to hours depending on the security level [7]. Their attack relied on embedding elliptic curves into abelian surfaces, and used Kani's reducibility criterion [18] as part of their decisional oracle. The attack got improved by a quick series of follow-up works using a direct computational approach [22,24], and finally Robert managed to prove that even if the endomorphism ring of the starting curve in SIDH is unknown, there is always a polynomial-time attack by using abelian eightfolds [25].Despite these generalizations and the increasing interest in higher-dimensional cryptographic applications, most of the aforementioned implementations restrict themselves to isogenies of very low prime degree. The genus-2 version of the CGL hash function in [8] used (2, 2)-isogenies only, since they are by far easiest to compute. Kunzweiler improved further on these (2, 2)-isogeny formulae in [19], and Castyck and Decru provided a (3, 3)-version based on their multiradical isogeny setting [6]. The (now also broken) genus-2 variant of SIDH in [14] used (2, 2)and (3, 3)-isogenies to obtain a five-minute key exchange on the basic security level. The implementation of the attacks on SIKE in [7] and [22] only target Bob's private key, since this requires only using (2, 2)-isogenies. In [26], Santos, Costello and Frengley do manage to use up to (11, 11)-isogenies, but only as a decisional tool to detect (N, N )-split Jacobians.The reason for these restrictions is that computing isogenies between abelian surfaces is typically a lot harder than isogenies between elliptic curves. The general (ℓ, ℓ)-isogeny formulae by Cosset and Robert [12] have polynomial time complexity O(ℓ 2 ) or O(ℓ 4 ), depending on ℓ mod 4, but arithmetic has to be performed in the field extension where the theta coordinates are defined, which can turn expensive quickly for cryptographic purposes. The only other known general parametrization are the (3, 3)-isogeny formulae by Bruin, Flynn and Testa (BFT) [4], which were used as a basis for both the multiradical (3, 3)-hash function and the genus-2 variant of SIDH. The parametrization is complete, but the formulae require over 37.500 multiplications if on...