"Padding is invalid and cannot be removed" using AesManaged

The trick is to use MemoryStream.ToArray(). I also changed your code so that it uses the CryptoStream to Write, in both encrypting and decrypting. And you don't need to call CryptoStream.FlushFinalBlock() explicitly, because you have it in a using() statement, and that flush will happen on Dispose(). The following works for me.

byte[] rawPlaintext = System.Text.Encoding.Unicode.GetBytes("This is all clear now!");

using (Aes aes = new AesManaged())
{
    aes.Padding = PaddingMode.PKCS7;
    aes.KeySize = 128;          // in bits
    aes.Key = new byte[128/8];  // 16 bytes for 128 bit encryption
    aes.IV = new byte[128/8];   // AES needs a 16-byte IV
    // Should set Key and IV here.  Good approach: derive them from 
    // a password via Cryptography.Rfc2898DeriveBytes 
    byte[] cipherText= null;
    byte[] plainText= null;

    using (MemoryStream ms = new MemoryStream())
    {
        using (CryptoStream cs = new CryptoStream(ms, aes.CreateEncryptor(), CryptoStreamMode.Write))
        {
            cs.Write(rawPlaintext, 0, rawPlaintext.Length);
        }

        cipherText= ms.ToArray();
    }


    using (MemoryStream ms = new MemoryStream())
    {
        using (CryptoStream cs = new CryptoStream(ms, aes.CreateDecryptor(), CryptoStreamMode.Write))
        {
            cs.Write(cipherText, 0, cipherText.Length);
        }

        plainText = ms.ToArray();
    }
    string s = System.Text.Encoding.Unicode.GetString(plainText);
    Console.WriteLine(s);
}

Also, I guess you know you will want to explicitly set the Mode of the AesManaged instance, and use System.Security.Cryptography.Rfc2898DeriveBytes to derive the Key and IV from a password and salt.

see also:
- AesManaged

This exception can be caused by a mismatch of any one of a number of encryption parameters.

I used the Security.Cryptography.Debug interface to trace all parameters used in the encrypt/decrypt methods.

Finally I found out that my problem was that I set the KeySize property after setting the Key causing the class to regenerate a random key and not using the key that I was initially set up.

For whats its worth, I'll document what I faced. I was trying to read the encryptor memory stream before the CryptoStream was closed. I was naive and I wasted a day debugging it.

    public static byte[] Encrypt(byte[] buffer, byte[] sessionKey, out byte[] iv)
    {
        byte[] encrypted;
        iv = null;
        using (AesCryptoServiceProvider aesAlg = new AesCryptoServiceProvider { Mode = CipherMode.CBC, Padding = PaddingMode.PKCS7 })
        {
            aesAlg.Key = sessionKey;
            iv = aesAlg.IV;
            ICryptoTransform encryptor = aesAlg.CreateEncryptor(sessionKey, iv);

            // Create the streams used for encryption.
            using (MemoryStream msEncrypt = new MemoryStream())
            {
                using (CryptoStream csEncrypt = new CryptoStream(msEncrypt, encryptor, CryptoStreamMode.Write))
                {
                    csEncrypt.Write(buffer, 0, buffer.Length);
                    
                    //This was not closing the cryptostream and only worked if I called FlushFinalBlock()
                    //encrypted = msEncrypt.ToArray(); 
                }

                encrypted = msEncrypt.ToArray();

                return encrypted;
            }
        }
    }

Moving the encryptor memory stream read after the cypto stream was closed solved the problem. As Cheeso mentioned. You don't need to call the FlushFinalBlock() if you're using the using block.

byte[] rawData = new byte[rawPlaintext.Length];

You need to read the length of the buffer, that probably includes the necessary padding (IIRC, been a few years).

Nobody answered, that actually MemoryStream.GetBuffer returns the allocated buffer, not the real data in this buffer. In this case it returns 256-byte buffer, while it contains only 32 bytes of encrypted data.